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"
13 import Type hiding ( substTy, extendTvSubst, substTyVar )
16 import FamInstEnv ( FamInstEnv )
18 import MkId ( seqId, realWorldPrimId )
19 import MkCore ( mkImpossibleExpr )
22 import Name ( mkSystemVarName, isExternalName )
24 import OptCoercion ( optCoercion )
25 import FamInstEnv ( topNormaliseType )
26 import DataCon ( DataCon, dataConWorkId, dataConRepStrictness )
27 import CoreMonad ( SimplifierSwitch(..), Tick(..) )
29 import Demand ( isStrictDmd, splitStrictSig )
30 import PprCore ( pprParendExpr, pprCoreExpr )
31 import CoreUnfold ( mkUnfolding, mkCoreUnfolding
32 , mkInlineUnfolding, mkSimpleUnfolding
33 , exprIsConApp_maybe, callSiteInline, CallCtxt(..) )
35 import qualified CoreSubst
36 import CoreArity ( exprArity )
37 import Rules ( lookupRule, getRules )
38 import BasicTypes ( isMarkedStrict, Arity )
39 import CostCentre ( currentCCS, pushCCisNop )
40 import TysPrim ( realWorldStatePrimTy )
41 import BasicTypes ( TopLevelFlag(..), isTopLevel, RecFlag(..) )
42 import MonadUtils ( foldlM, mapAccumLM )
43 import Maybes ( orElse )
44 import Data.List ( mapAccumL )
50 The guts of the simplifier is in this module, but the driver loop for
51 the simplifier is in SimplCore.lhs.
54 -----------------------------------------
55 *** IMPORTANT NOTE ***
56 -----------------------------------------
57 The simplifier used to guarantee that the output had no shadowing, but
58 it does not do so any more. (Actually, it never did!) The reason is
59 documented with simplifyArgs.
62 -----------------------------------------
63 *** IMPORTANT NOTE ***
64 -----------------------------------------
65 Many parts of the simplifier return a bunch of "floats" as well as an
66 expression. This is wrapped as a datatype SimplUtils.FloatsWith.
68 All "floats" are let-binds, not case-binds, but some non-rec lets may
69 be unlifted (with RHS ok-for-speculation).
73 -----------------------------------------
74 ORGANISATION OF FUNCTIONS
75 -----------------------------------------
77 - simplify all top-level binders
78 - for NonRec, call simplRecOrTopPair
79 - for Rec, call simplRecBind
82 ------------------------------
83 simplExpr (applied lambda) ==> simplNonRecBind
84 simplExpr (Let (NonRec ...) ..) ==> simplNonRecBind
85 simplExpr (Let (Rec ...) ..) ==> simplify binders; simplRecBind
87 ------------------------------
88 simplRecBind [binders already simplfied]
89 - use simplRecOrTopPair on each pair in turn
91 simplRecOrTopPair [binder already simplified]
92 Used for: recursive bindings (top level and nested)
93 top-level non-recursive bindings
95 - check for PreInlineUnconditionally
99 Used for: non-top-level non-recursive bindings
100 beta reductions (which amount to the same thing)
101 Because it can deal with strict arts, it takes a
102 "thing-inside" and returns an expression
104 - check for PreInlineUnconditionally
105 - simplify binder, including its IdInfo
114 simplNonRecX: [given a *simplified* RHS, but an *unsimplified* binder]
115 Used for: binding case-binder and constr args in a known-constructor case
116 - check for PreInLineUnconditionally
120 ------------------------------
121 simplLazyBind: [binder already simplified, RHS not]
122 Used for: recursive bindings (top level and nested)
123 top-level non-recursive bindings
124 non-top-level, but *lazy* non-recursive bindings
125 [must not be strict or unboxed]
126 Returns floats + an augmented environment, not an expression
127 - substituteIdInfo and add result to in-scope
128 [so that rules are available in rec rhs]
131 - float if exposes constructor or PAP
135 completeNonRecX: [binder and rhs both simplified]
136 - if the the thing needs case binding (unlifted and not ok-for-spec)
142 completeBind: [given a simplified RHS]
143 [used for both rec and non-rec bindings, top level and not]
144 - try PostInlineUnconditionally
145 - add unfolding [this is the only place we add an unfolding]
150 Right hand sides and arguments
151 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
152 In many ways we want to treat
153 (a) the right hand side of a let(rec), and
154 (b) a function argument
155 in the same way. But not always! In particular, we would
156 like to leave these arguments exactly as they are, so they
157 will match a RULE more easily.
162 It's harder to make the rule match if we ANF-ise the constructor,
163 or eta-expand the PAP:
165 f (let { a = g x; b = h x } in (a,b))
168 On the other hand if we see the let-defns
173 then we *do* want to ANF-ise and eta-expand, so that p and q
174 can be safely inlined.
176 Even floating lets out is a bit dubious. For let RHS's we float lets
177 out if that exposes a value, so that the value can be inlined more vigorously.
180 r = let x = e in (x,x)
182 Here, if we float the let out we'll expose a nice constructor. We did experiments
183 that showed this to be a generally good thing. But it was a bad thing to float
184 lets out unconditionally, because that meant they got allocated more often.
186 For function arguments, there's less reason to expose a constructor (it won't
187 get inlined). Just possibly it might make a rule match, but I'm pretty skeptical.
188 So for the moment we don't float lets out of function arguments either.
193 For eta expansion, we want to catch things like
195 case e of (a,b) -> \x -> case a of (p,q) -> \y -> r
197 If the \x was on the RHS of a let, we'd eta expand to bring the two
198 lambdas together. And in general that's a good thing to do. Perhaps
199 we should eta expand wherever we find a (value) lambda? Then the eta
200 expansion at a let RHS can concentrate solely on the PAP case.
203 %************************************************************************
205 \subsection{Bindings}
207 %************************************************************************
210 simplTopBinds :: SimplEnv -> [InBind] -> SimplM SimplEnv
212 simplTopBinds env0 binds0
213 = do { -- Put all the top-level binders into scope at the start
214 -- so that if a transformation rule has unexpectedly brought
215 -- anything into scope, then we don't get a complaint about that.
216 -- It's rather as if the top-level binders were imported.
217 ; env1 <- simplRecBndrs env0 (bindersOfBinds binds0)
218 ; dflags <- getDOptsSmpl
219 ; let dump_flag = dopt Opt_D_verbose_core2core dflags
220 ; env2 <- simpl_binds dump_flag env1 binds0
221 ; freeTick SimplifierDone
224 -- We need to track the zapped top-level binders, because
225 -- they should have their fragile IdInfo zapped (notably occurrence info)
226 -- That's why we run down binds and bndrs' simultaneously.
228 -- The dump-flag emits a trace for each top-level binding, which
229 -- helps to locate the tracing for inlining and rule firing
230 simpl_binds :: Bool -> SimplEnv -> [InBind] -> SimplM SimplEnv
231 simpl_binds _ env [] = return env
232 simpl_binds dump env (bind:binds) = do { env' <- trace_bind dump bind $
234 ; simpl_binds dump env' binds }
236 trace_bind True bind = pprTrace "SimplBind" (ppr (bindersOf bind))
237 trace_bind False _ = \x -> x
239 simpl_bind env (Rec pairs) = simplRecBind env TopLevel pairs
240 simpl_bind env (NonRec b r) = simplRecOrTopPair env' TopLevel b b' r
242 (env', b') = addBndrRules env b (lookupRecBndr env b)
246 %************************************************************************
248 \subsection{Lazy bindings}
250 %************************************************************************
252 simplRecBind is used for
253 * recursive bindings only
256 simplRecBind :: SimplEnv -> TopLevelFlag
259 simplRecBind env0 top_lvl pairs0
260 = do { let (env_with_info, triples) = mapAccumL add_rules env0 pairs0
261 ; env1 <- go (zapFloats env_with_info) triples
262 ; return (env0 `addRecFloats` env1) }
263 -- addFloats adds the floats from env1,
264 -- _and_ updates env0 with the in-scope set from env1
266 add_rules :: SimplEnv -> (InBndr,InExpr) -> (SimplEnv, (InBndr, OutBndr, InExpr))
267 -- Add the (substituted) rules to the binder
268 add_rules env (bndr, rhs) = (env', (bndr, bndr', rhs))
270 (env', bndr') = addBndrRules env bndr (lookupRecBndr env bndr)
272 go env [] = return env
274 go env ((old_bndr, new_bndr, rhs) : pairs)
275 = do { env' <- simplRecOrTopPair env top_lvl old_bndr new_bndr rhs
279 simplOrTopPair is used for
280 * recursive bindings (whether top level or not)
281 * top-level non-recursive bindings
283 It assumes the binder has already been simplified, but not its IdInfo.
286 simplRecOrTopPair :: SimplEnv
288 -> InId -> OutBndr -> InExpr -- Binder and rhs
289 -> SimplM SimplEnv -- Returns an env that includes the binding
291 simplRecOrTopPair env top_lvl old_bndr new_bndr rhs
292 | preInlineUnconditionally env top_lvl old_bndr rhs -- Check for unconditional inline
293 = do { tick (PreInlineUnconditionally old_bndr)
294 ; return (extendIdSubst env old_bndr (mkContEx env rhs)) }
297 = simplLazyBind env top_lvl Recursive old_bndr new_bndr rhs env
298 -- May not actually be recursive, but it doesn't matter
302 simplLazyBind is used for
303 * [simplRecOrTopPair] recursive bindings (whether top level or not)
304 * [simplRecOrTopPair] top-level non-recursive bindings
305 * [simplNonRecE] non-top-level *lazy* non-recursive bindings
308 1. It assumes that the binder is *already* simplified,
309 and is in scope, and its IdInfo too, except unfolding
311 2. It assumes that the binder type is lifted.
313 3. It does not check for pre-inline-unconditionallly;
314 that should have been done already.
317 simplLazyBind :: SimplEnv
318 -> TopLevelFlag -> RecFlag
319 -> InId -> OutId -- Binder, both pre-and post simpl
320 -- The OutId has IdInfo, except arity, unfolding
321 -> InExpr -> SimplEnv -- The RHS and its environment
324 simplLazyBind env top_lvl is_rec bndr bndr1 rhs rhs_se
325 = -- pprTrace "simplLazyBind" ((ppr bndr <+> ppr bndr1) $$ ppr rhs $$ ppr (seIdSubst rhs_se)) $
326 do { let rhs_env = rhs_se `setInScope` env
327 (tvs, body) = case collectTyBinders rhs of
328 (tvs, body) | not_lam body -> (tvs,body)
329 | otherwise -> ([], rhs)
330 not_lam (Lam _ _) = False
332 -- Do not do the "abstract tyyvar" thing if there's
333 -- a lambda inside, becuase it defeats eta-reduction
334 -- f = /\a. \x. g a x
337 ; (body_env, tvs') <- simplBinders rhs_env tvs
338 -- See Note [Floating and type abstraction] in SimplUtils
341 ; (body_env1, body1) <- simplExprF body_env body mkRhsStop
342 -- ANF-ise a constructor or PAP rhs
343 ; (body_env2, body2) <- prepareRhs top_lvl body_env1 bndr1 body1
346 <- if not (doFloatFromRhs top_lvl is_rec False body2 body_env2)
347 then -- No floating, revert to body1
348 do { rhs' <- mkLam env tvs' (wrapFloats body_env1 body1)
349 ; return (env, rhs') }
351 else if null tvs then -- Simple floating
352 do { tick LetFloatFromLet
353 ; return (addFloats env body_env2, body2) }
355 else -- Do type-abstraction first
356 do { tick LetFloatFromLet
357 ; (poly_binds, body3) <- abstractFloats tvs' body_env2 body2
358 ; rhs' <- mkLam env tvs' body3
359 ; env' <- foldlM (addPolyBind top_lvl) env poly_binds
360 ; return (env', rhs') }
362 ; completeBind env' top_lvl bndr bndr1 rhs' }
365 A specialised variant of simplNonRec used when the RHS is already simplified,
366 notably in knownCon. It uses case-binding where necessary.
369 simplNonRecX :: SimplEnv
370 -> InId -- Old binder
371 -> OutExpr -- Simplified RHS
374 simplNonRecX env bndr new_rhs
375 | isDeadBinder bndr -- Not uncommon; e.g. case (a,b) of b { (p,q) -> p }
376 = return env -- Here b is dead, and we avoid creating
377 | otherwise -- the binding b = (a,b)
378 = do { (env', bndr') <- simplBinder env bndr
379 ; completeNonRecX NotTopLevel env' (isStrictId bndr) bndr bndr' new_rhs }
380 -- simplNonRecX is only used for NotTopLevel things
382 completeNonRecX :: TopLevelFlag -> SimplEnv
384 -> InId -- Old binder
385 -> OutId -- New binder
386 -> OutExpr -- Simplified RHS
389 completeNonRecX top_lvl env is_strict old_bndr new_bndr new_rhs
390 = do { (env1, rhs1) <- prepareRhs top_lvl (zapFloats env) new_bndr new_rhs
392 if doFloatFromRhs NotTopLevel NonRecursive is_strict rhs1 env1
393 then do { tick LetFloatFromLet
394 ; return (addFloats env env1, rhs1) } -- Add the floats to the main env
395 else return (env, wrapFloats env1 rhs1) -- Wrap the floats around the RHS
396 ; completeBind env2 NotTopLevel old_bndr new_bndr rhs2 }
399 {- No, no, no! Do not try preInlineUnconditionally in completeNonRecX
400 Doing so risks exponential behaviour, because new_rhs has been simplified once already
401 In the cases described by the folowing commment, postInlineUnconditionally will
402 catch many of the relevant cases.
403 -- This happens; for example, the case_bndr during case of
404 -- known constructor: case (a,b) of x { (p,q) -> ... }
405 -- Here x isn't mentioned in the RHS, so we don't want to
406 -- create the (dead) let-binding let x = (a,b) in ...
408 -- Similarly, single occurrences can be inlined vigourously
409 -- e.g. case (f x, g y) of (a,b) -> ....
410 -- If a,b occur once we can avoid constructing the let binding for them.
412 Furthermore in the case-binding case preInlineUnconditionally risks extra thunks
413 -- Consider case I# (quotInt# x y) of
414 -- I# v -> let w = J# v in ...
415 -- If we gaily inline (quotInt# x y) for v, we end up building an
417 -- let w = J# (quotInt# x y) in ...
418 -- because quotInt# can fail.
420 | preInlineUnconditionally env NotTopLevel bndr new_rhs
421 = thing_inside (extendIdSubst env bndr (DoneEx new_rhs))
424 ----------------------------------
425 prepareRhs takes a putative RHS, checks whether it's a PAP or
426 constructor application and, if so, converts it to ANF, so that the
427 resulting thing can be inlined more easily. Thus
434 We also want to deal well cases like this
435 v = (f e1 `cast` co) e2
436 Here we want to make e1,e2 trivial and get
437 x1 = e1; x2 = e2; v = (f x1 `cast` co) v2
438 That's what the 'go' loop in prepareRhs does
441 prepareRhs :: TopLevelFlag -> SimplEnv -> OutId -> OutExpr -> SimplM (SimplEnv, OutExpr)
442 -- Adds new floats to the env iff that allows us to return a good RHS
443 prepareRhs top_lvl env id (Cast rhs co) -- Note [Float coercions]
444 | (ty1, _ty2) <- coercionKind co -- Do *not* do this if rhs has an unlifted type
445 , not (isUnLiftedType ty1) -- see Note [Float coercions (unlifted)]
446 = do { (env', rhs') <- makeTrivialWithInfo top_lvl env sanitised_info rhs
447 ; return (env', Cast rhs' co) }
449 sanitised_info = vanillaIdInfo `setStrictnessInfo` strictnessInfo info
450 `setDemandInfo` demandInfo info
453 prepareRhs top_lvl env0 _ rhs0
454 = do { (_is_exp, env1, rhs1) <- go 0 env0 rhs0
455 ; return (env1, rhs1) }
457 go n_val_args env (Cast rhs co)
458 = do { (is_exp, env', rhs') <- go n_val_args env rhs
459 ; return (is_exp, env', Cast rhs' co) }
460 go n_val_args env (App fun (Type ty))
461 = do { (is_exp, env', rhs') <- go n_val_args env fun
462 ; return (is_exp, env', App rhs' (Type ty)) }
463 go n_val_args env (App fun arg)
464 = do { (is_exp, env', fun') <- go (n_val_args+1) env fun
466 True -> do { (env'', arg') <- makeTrivial top_lvl env' arg
467 ; return (True, env'', App fun' arg') }
468 False -> return (False, env, App fun arg) }
469 go n_val_args env (Var fun)
470 = return (is_exp, env, Var fun)
472 is_exp = isExpandableApp fun n_val_args -- The fun a constructor or PAP
473 -- See Note [CONLIKE pragma] in BasicTypes
474 -- The definition of is_exp should match that in
475 -- OccurAnal.occAnalApp
478 = return (False, env, other)
482 Note [Float coercions]
483 ~~~~~~~~~~~~~~~~~~~~~~
484 When we find the binding
486 we'd like to transform it to
488 x = x `cast` co -- A trivial binding
489 There's a chance that e will be a constructor application or function, or something
490 like that, so moving the coerion to the usage site may well cancel the coersions
491 and lead to further optimisation. Example:
494 data instance T Int = T Int
496 foo :: Int -> Int -> Int
501 go n = case x of { T m -> go (n-m) }
502 -- This case should optimise
504 Note [Preserve strictness when floating coercions]
505 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
506 In the Note [Float coercions] transformation, keep the strictness info.
508 f = e `cast` co -- f has strictness SSL
510 f' = e -- f' also has strictness SSL
511 f = f' `cast` co -- f still has strictness SSL
513 Its not wrong to drop it on the floor, but better to keep it.
515 Note [Float coercions (unlifted)]
516 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
517 BUT don't do [Float coercions] if 'e' has an unlifted type.
520 foo :: Int = (error (# Int,Int #) "urk")
521 `cast` CoUnsafe (# Int,Int #) Int
523 If do the makeTrivial thing to the error call, we'll get
524 foo = case error (# Int,Int #) "urk" of v -> v `cast` ...
525 But 'v' isn't in scope!
527 These strange casts can happen as a result of case-of-case
528 bar = case (case x of { T -> (# 2,3 #); F -> error "urk" }) of
533 makeTrivial :: TopLevelFlag -> SimplEnv -> OutExpr -> SimplM (SimplEnv, OutExpr)
534 -- Binds the expression to a variable, if it's not trivial, returning the variable
535 makeTrivial top_lvl env expr = makeTrivialWithInfo top_lvl env vanillaIdInfo expr
537 makeTrivialWithInfo :: TopLevelFlag -> SimplEnv -> IdInfo
538 -> OutExpr -> SimplM (SimplEnv, OutExpr)
539 -- Propagate strictness and demand info to the new binder
540 -- Note [Preserve strictness when floating coercions]
541 -- Returned SimplEnv has same substitution as incoming one
542 makeTrivialWithInfo top_lvl env info expr
543 | exprIsTrivial expr -- Already trivial
544 || not (bindingOk top_lvl expr expr_ty) -- Cannot trivialise
545 -- See Note [Cannot trivialise]
547 | otherwise -- See Note [Take care] below
548 = do { uniq <- getUniqueM
549 ; let name = mkSystemVarName uniq (fsLit "a")
550 var = mkLocalIdWithInfo name expr_ty info
551 ; env' <- completeNonRecX top_lvl env False var var expr
552 ; expr' <- simplVar env' var
553 ; return (env', expr') }
554 -- The simplVar is needed becase we're constructing a new binding
556 -- And if rhs is of form (rhs1 |> co), then we might get
559 -- and now a's RHS is trivial and can be substituted out, and that
560 -- is what completeNonRecX will do
561 -- To put it another way, it's as if we'd simplified
562 -- let var = e in var
564 expr_ty = exprType expr
566 bindingOk :: TopLevelFlag -> CoreExpr -> Type -> Bool
567 -- True iff we can have a binding of this expression at this level
568 -- Precondition: the type is the type of the expression
569 bindingOk top_lvl _ expr_ty
570 | isTopLevel top_lvl = not (isUnLiftedType expr_ty)
574 Note [Cannot trivialise]
575 ~~~~~~~~~~~~~~~~~~~~~~~~
582 Then we can't ANF-ise foo, even though we'd like to, because
583 we can't make a top-level binding for the Addr# (f 3). And if
584 so we don't want to turn it into
585 foo = let x = f 3 in Bar x
586 because we'll just end up inlining x back, and that makes the
587 simplifier loop. Better not to ANF-ise it at all.
589 A case in point is literal strings (a MachStr is not regarded as
594 We don't want to ANF-ise this.
596 %************************************************************************
598 \subsection{Completing a lazy binding}
600 %************************************************************************
603 * deals only with Ids, not TyVars
604 * takes an already-simplified binder and RHS
605 * is used for both recursive and non-recursive bindings
606 * is used for both top-level and non-top-level bindings
608 It does the following:
609 - tries discarding a dead binding
610 - tries PostInlineUnconditionally
611 - add unfolding [this is the only place we add an unfolding]
614 It does *not* attempt to do let-to-case. Why? Because it is used for
615 - top-level bindings (when let-to-case is impossible)
616 - many situations where the "rhs" is known to be a WHNF
617 (so let-to-case is inappropriate).
619 Nor does it do the atomic-argument thing
622 completeBind :: SimplEnv
623 -> TopLevelFlag -- Flag stuck into unfolding
624 -> InId -- Old binder
625 -> OutId -> OutExpr -- New binder and RHS
627 -- completeBind may choose to do its work
628 -- * by extending the substitution (e.g. let x = y in ...)
629 -- * or by adding to the floats in the envt
631 completeBind env top_lvl old_bndr new_bndr new_rhs
632 = do { let old_info = idInfo old_bndr
633 old_unf = unfoldingInfo old_info
634 occ_info = occInfo old_info
636 ; new_unfolding <- simplUnfolding env top_lvl old_bndr occ_info new_rhs old_unf
638 ; if postInlineUnconditionally env top_lvl new_bndr occ_info new_rhs new_unfolding
639 -- Inline and discard the binding
640 then do { tick (PostInlineUnconditionally old_bndr)
641 ; -- pprTrace "postInlineUnconditionally" (ppr old_bndr <+> equals <+> ppr new_rhs) $
642 return (extendIdSubst env old_bndr (DoneEx new_rhs)) }
643 -- Use the substitution to make quite, quite sure that the
644 -- substitution will happen, since we are going to discard the binding
646 else return (addNonRecWithUnf env new_bndr new_rhs new_unfolding) }
648 ------------------------------
649 addPolyBind :: TopLevelFlag -> SimplEnv -> OutBind -> SimplM SimplEnv
650 -- Add a new binding to the environment, complete with its unfolding
651 -- but *do not* do postInlineUnconditionally, because we have already
652 -- processed some of the scope of the binding
653 -- We still want the unfolding though. Consider
655 -- x = /\a. let y = ... in Just y
657 -- Then we float the y-binding out (via abstractFloats and addPolyBind)
658 -- but 'x' may well then be inlined in 'body' in which case we'd like the
659 -- opportunity to inline 'y' too.
661 addPolyBind top_lvl env (NonRec poly_id rhs)
662 = do { unfolding <- simplUnfolding env top_lvl poly_id NoOccInfo rhs noUnfolding
663 -- Assumes that poly_id did not have an INLINE prag
664 -- which is perhaps wrong. ToDo: think about this
665 ; return (addNonRecWithUnf env poly_id rhs unfolding) }
667 addPolyBind _ env bind@(Rec _) = return (extendFloats env bind)
668 -- Hack: letrecs are more awkward, so we extend "by steam"
669 -- without adding unfoldings etc. At worst this leads to
670 -- more simplifier iterations
672 ------------------------------
673 addNonRecWithUnf :: SimplEnv
674 -> OutId -> OutExpr -- New binder and RHS
675 -> Unfolding -- New unfolding
677 addNonRecWithUnf env new_bndr new_rhs new_unfolding
678 = let new_arity = exprArity new_rhs
679 old_arity = idArity new_bndr
680 info1 = idInfo new_bndr `setArityInfo` new_arity
682 -- Unfolding info: Note [Setting the new unfolding]
683 info2 = info1 `setUnfoldingInfo` new_unfolding
685 -- Demand info: Note [Setting the demand info]
686 info3 | isEvaldUnfolding new_unfolding = zapDemandInfo info2 `orElse` info2
689 final_id = new_bndr `setIdInfo` info3
690 dmd_arity = length $ fst $ splitStrictSig $ idStrictness new_bndr
692 ASSERT( isId new_bndr )
693 WARN( new_arity < old_arity || new_arity < dmd_arity,
694 (ptext (sLit "Arity decrease:") <+> (ppr final_id <+> ppr old_arity
695 <+> ppr new_arity <+> ppr dmd_arity) $$ ppr new_rhs) )
696 -- Note [Arity decrease]
698 final_id `seq` -- This seq forces the Id, and hence its IdInfo,
699 -- and hence any inner substitutions
700 -- pprTrace "Binding" (ppr final_id <+> ppr unfolding) $
701 addNonRec env final_id new_rhs
702 -- The addNonRec adds it to the in-scope set too
704 ------------------------------
705 simplUnfolding :: SimplEnv-> TopLevelFlag
707 -> OccInfo -> OutExpr
708 -> Unfolding -> SimplM Unfolding
709 -- Note [Setting the new unfolding]
710 simplUnfolding env _ _ _ _ (DFunUnfolding ar con ops)
711 = return (DFunUnfolding ar con ops')
713 ops' = map (substExpr (text "simplUnfolding") env) ops
715 simplUnfolding env top_lvl id _ _
716 (CoreUnfolding { uf_tmpl = expr, uf_arity = arity
717 , uf_src = src, uf_guidance = guide })
719 = do { expr' <- simplExpr rule_env expr
720 ; let src' = CoreSubst.substUnfoldingSource (mkCoreSubst (text "inline-unf") env) src
721 ; return (mkCoreUnfolding (isTopLevel top_lvl) src' expr' arity guide) }
722 -- See Note [Top-level flag on inline rules] in CoreUnfold
724 act = idInlineActivation id
725 rule_env = updMode (updModeForInlineRules act) env
726 -- See Note [Simplifying inside InlineRules] in SimplUtils
728 simplUnfolding _ top_lvl id _occ_info new_rhs _
729 = return (mkUnfolding InlineRhs (isTopLevel top_lvl) (isBottomingId id) new_rhs)
730 -- We make an unfolding *even for loop-breakers*.
731 -- Reason: (a) It might be useful to know that they are WHNF
732 -- (b) In TidyPgm we currently assume that, if we want to
733 -- expose the unfolding then indeed we *have* an unfolding
734 -- to expose. (We could instead use the RHS, but currently
735 -- we don't.) The simple thing is always to have one.
738 Note [Arity decrease]
739 ~~~~~~~~~~~~~~~~~~~~~
740 Generally speaking the arity of a binding should not decrease. But it *can*
741 legitimately happen becuase of RULES. Eg
743 where g has arity 2, will have arity 2. But if there's a rewrite rule
745 where h has arity 1, then f's arity will decrease. Here's a real-life example,
746 which is in the output of Specialise:
749 $dm {Arity 2} = \d.\x. op d
750 {-# RULES forall d. $dm Int d = $s$dm #-}
752 dInt = MkD .... opInt ...
753 opInt {Arity 1} = $dm dInt
755 $s$dm {Arity 0} = \x. op dInt }
757 Here opInt has arity 1; but when we apply the rule its arity drops to 0.
758 That's why Specialise goes to a little trouble to pin the right arity
759 on specialised functions too.
761 Note [Setting the new unfolding]
762 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
763 * If there's an INLINE pragma, we simplify the RHS gently. Maybe we
764 should do nothing at all, but simplifying gently might get rid of
767 * If not, we make an unfolding from the new RHS. But *only* for
768 non-loop-breakers. Making loop breakers not have an unfolding at all
769 means that we can avoid tests in exprIsConApp, for example. This is
770 important: if exprIsConApp says 'yes' for a recursive thing, then we
771 can get into an infinite loop
773 If there's an InlineRule on a loop breaker, we hang on to the inlining.
774 It's pretty dodgy, but the user did say 'INLINE'. May need to revisit
777 Note [Setting the demand info]
778 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
779 If the unfolding is a value, the demand info may
780 go pear-shaped, so we nuke it. Example:
782 case x of (p,q) -> h p q x
783 Here x is certainly demanded. But after we've nuked
784 the case, we'll get just
785 let x = (a,b) in h a b x
786 and now x is not demanded (I'm assuming h is lazy)
787 This really happens. Similarly
788 let f = \x -> e in ...f..f...
789 After inlining f at some of its call sites the original binding may
790 (for example) be no longer strictly demanded.
791 The solution here is a bit ad hoc...
794 %************************************************************************
796 \subsection[Simplify-simplExpr]{The main function: simplExpr}
798 %************************************************************************
800 The reason for this OutExprStuff stuff is that we want to float *after*
801 simplifying a RHS, not before. If we do so naively we get quadratic
802 behaviour as things float out.
804 To see why it's important to do it after, consider this (real) example:
818 a -- Can't inline a this round, cos it appears twice
822 Each of the ==> steps is a round of simplification. We'd save a
823 whole round if we float first. This can cascade. Consider
828 let f = let d1 = ..d.. in \y -> e
832 in \x -> ...(\y ->e)...
834 Only in this second round can the \y be applied, and it
835 might do the same again.
839 simplExpr :: SimplEnv -> CoreExpr -> SimplM CoreExpr
840 simplExpr env expr = simplExprC env expr mkBoringStop
842 simplExprC :: SimplEnv -> CoreExpr -> SimplCont -> SimplM CoreExpr
843 -- Simplify an expression, given a continuation
844 simplExprC env expr cont
845 = -- pprTrace "simplExprC" (ppr expr $$ ppr cont {- $$ ppr (seIdSubst env) -} $$ ppr (seFloats env) ) $
846 do { (env', expr') <- simplExprF (zapFloats env) expr cont
847 ; -- pprTrace "simplExprC ret" (ppr expr $$ ppr expr') $
848 -- pprTrace "simplExprC ret3" (ppr (seInScope env')) $
849 -- pprTrace "simplExprC ret4" (ppr (seFloats env')) $
850 return (wrapFloats env' expr') }
852 --------------------------------------------------
853 simplExprF :: SimplEnv -> InExpr -> SimplCont
854 -> SimplM (SimplEnv, OutExpr)
856 simplExprF env e cont
857 = -- pprTrace "simplExprF" (ppr e $$ ppr cont $$ ppr (seTvSubst env) $$ ppr (seIdSubst env) {- $$ ppr (seFloats env) -} ) $
858 simplExprF' env e cont
860 simplExprF' :: SimplEnv -> InExpr -> SimplCont
861 -> SimplM (SimplEnv, OutExpr)
862 simplExprF' env (Var v) cont = simplVarF env v cont
863 simplExprF' env (Lit lit) cont = rebuild env (Lit lit) cont
864 simplExprF' env (Note n expr) cont = simplNote env n expr cont
865 simplExprF' env (Cast body co) cont = simplCast env body co cont
866 simplExprF' env (App fun arg) cont = simplExprF env fun $
867 ApplyTo NoDup arg env cont
869 simplExprF' env expr@(Lam _ _) cont
870 = simplLam env (map zap bndrs) body cont
871 -- The main issue here is under-saturated lambdas
872 -- (\x1. \x2. e) arg1
873 -- Here x1 might have "occurs-once" occ-info, because occ-info
874 -- is computed assuming that a group of lambdas is applied
875 -- all at once. If there are too few args, we must zap the
878 n_args = countArgs cont
879 n_params = length bndrs
880 (bndrs, body) = collectBinders expr
881 zap | n_args >= n_params = \b -> b
882 | otherwise = \b -> if isTyCoVar b then b
884 -- NB: we count all the args incl type args
885 -- so we must count all the binders (incl type lambdas)
887 simplExprF' env (Type ty) cont
888 = ASSERT( contIsRhsOrArg cont )
889 do { ty' <- simplCoercion env ty
890 ; rebuild env (Type ty') cont }
892 simplExprF' env (Case scrut bndr _ alts) cont
893 | not (switchIsOn (getSwitchChecker env) NoCaseOfCase)
894 = -- Simplify the scrutinee with a Select continuation
895 simplExprF env scrut (Select NoDup bndr alts env cont)
898 = -- If case-of-case is off, simply simplify the case expression
899 -- in a vanilla Stop context, and rebuild the result around it
900 do { case_expr' <- simplExprC env scrut
901 (Select NoDup bndr alts env mkBoringStop)
902 ; rebuild env case_expr' cont }
904 simplExprF' env (Let (Rec pairs) body) cont
905 = do { env' <- simplRecBndrs env (map fst pairs)
906 -- NB: bndrs' don't have unfoldings or rules
907 -- We add them as we go down
909 ; env'' <- simplRecBind env' NotTopLevel pairs
910 ; simplExprF env'' body cont }
912 simplExprF' env (Let (NonRec bndr rhs) body) cont
913 = simplNonRecE env bndr (rhs, env) ([], body) cont
915 ---------------------------------
916 simplType :: SimplEnv -> InType -> SimplM OutType
917 -- Kept monadic just so we can do the seqType
919 = -- pprTrace "simplType" (ppr ty $$ ppr (seTvSubst env)) $
920 seqType new_ty `seq` return new_ty
922 new_ty = substTy env ty
924 ---------------------------------
925 simplCoercion :: SimplEnv -> InType -> SimplM OutType
926 -- The InType isn't *necessarily* a coercion, but it might be
927 -- (in a type application, say) and optCoercion is a no-op on types
929 = seqType new_co `seq` return new_co
931 new_co = optCoercion (getTvSubst env) co
935 %************************************************************************
937 \subsection{The main rebuilder}
939 %************************************************************************
942 rebuild :: SimplEnv -> OutExpr -> SimplCont -> SimplM (SimplEnv, OutExpr)
943 -- At this point the substitution in the SimplEnv should be irrelevant
944 -- only the in-scope set and floats should matter
945 rebuild env expr cont0
946 = -- pprTrace "rebuild" (ppr expr $$ ppr cont0 $$ ppr (seFloats env)) $
948 Stop {} -> return (env, expr)
949 CoerceIt co cont -> rebuild env (mkCoerce co expr) cont
950 Select _ bndr alts se cont -> rebuildCase (se `setFloats` env) expr bndr alts cont
951 StrictArg info _ cont -> rebuildCall env (info `addArgTo` expr) cont
952 StrictBind b bs body se cont -> do { env' <- simplNonRecX (se `setFloats` env) b expr
953 ; simplLam env' bs body cont }
954 ApplyTo dup_flag arg se cont -- See Note [Avoid redundant simplification]
955 | isSimplified dup_flag -> rebuild env (App expr arg) cont
956 | otherwise -> do { arg' <- simplExpr (se `setInScope` env) arg
957 ; rebuild env (App expr arg') cont }
961 %************************************************************************
965 %************************************************************************
968 simplCast :: SimplEnv -> InExpr -> Coercion -> SimplCont
969 -> SimplM (SimplEnv, OutExpr)
970 simplCast env body co0 cont0
971 = do { co1 <- simplCoercion env co0
972 ; simplExprF env body (addCoerce co1 cont0) }
974 addCoerce co cont = add_coerce co (coercionKind co) cont
976 add_coerce _co (s1, k1) cont -- co :: ty~ty
977 | s1 `coreEqType` k1 = cont -- is a no-op
979 add_coerce co1 (s1, _k2) (CoerceIt co2 cont)
980 | (_l1, t1) <- coercionKind co2
981 -- e |> (g1 :: S1~L) |> (g2 :: L~T1)
984 -- e |> (g1 . g2 :: S1~T1) otherwise
986 -- For example, in the initial form of a worker
987 -- we may find (coerce T (coerce S (\x.e))) y
988 -- and we'd like it to simplify to e[y/x] in one round
990 , s1 `coreEqType` t1 = cont -- The coerces cancel out
991 | otherwise = CoerceIt (mkTransCoercion co1 co2) cont
993 add_coerce co (s1s2, _t1t2) (ApplyTo dup (Type arg_ty) arg_se cont)
994 -- (f |> g) ty ---> (f ty) |> (g @ ty)
995 -- This implements the PushT and PushC rules from the paper
996 | Just (tyvar,_) <- splitForAllTy_maybe s1s2
998 (new_arg_ty, new_cast)
999 | isCoVar tyvar = (new_arg_co, mkCselRCoercion co) -- PushC rule
1000 | otherwise = (ty', mkInstCoercion co ty') -- PushT rule
1002 ApplyTo dup (Type new_arg_ty) (zapSubstEnv arg_se) (addCoerce new_cast cont)
1004 ty' = substTy (arg_se `setInScope` env) arg_ty
1005 new_arg_co = mkCsel1Coercion co `mkTransCoercion`
1006 ty' `mkTransCoercion`
1007 mkSymCoercion (mkCsel2Coercion co)
1009 add_coerce co (s1s2, _t1t2) (ApplyTo dup arg arg_se cont)
1010 | not (isTypeArg arg) -- This implements the Push rule from the paper
1011 , isFunTy s1s2 -- t1t2 must be a function type, becuase it's applied
1012 -- (e |> (g :: s1s2 ~ t1->t2)) f
1014 -- (e (f |> (arg g :: t1~s1))
1015 -- |> (res g :: s2->t2)
1017 -- t1t2 must be a function type, t1->t2, because it's applied
1018 -- to something but s1s2 might conceivably not be
1020 -- When we build the ApplyTo we can't mix the out-types
1021 -- with the InExpr in the argument, so we simply substitute
1022 -- to make it all consistent. It's a bit messy.
1023 -- But it isn't a common case.
1025 -- Example of use: Trac #995
1026 = ApplyTo dup new_arg (zapSubstEnv arg_se) (addCoerce co2 cont)
1028 -- we split coercion t1->t2 ~ s1->s2 into t1 ~ s1 and
1029 -- t2 ~ s2 with left and right on the curried form:
1030 -- (->) t1 t2 ~ (->) s1 s2
1031 [co1, co2] = decomposeCo 2 co
1032 new_arg = mkCoerce (mkSymCoercion co1) arg'
1033 arg' = substExpr (text "move-cast") (arg_se `setInScope` env) arg
1035 add_coerce co _ cont = CoerceIt co cont
1039 %************************************************************************
1041 \subsection{Lambdas}
1043 %************************************************************************
1046 simplLam :: SimplEnv -> [InId] -> InExpr -> SimplCont
1047 -> SimplM (SimplEnv, OutExpr)
1049 simplLam env [] body cont = simplExprF env body cont
1052 simplLam env (bndr:bndrs) body (ApplyTo _ arg arg_se cont)
1053 = do { tick (BetaReduction bndr)
1054 ; simplNonRecE env bndr (arg, arg_se) (bndrs, body) cont }
1056 -- Not enough args, so there are real lambdas left to put in the result
1057 simplLam env bndrs body cont
1058 = do { (env', bndrs') <- simplLamBndrs env bndrs
1059 ; body' <- simplExpr env' body
1060 ; new_lam <- mkLam env' bndrs' body'
1061 ; rebuild env' new_lam cont }
1064 simplNonRecE :: SimplEnv
1065 -> InBndr -- The binder
1066 -> (InExpr, SimplEnv) -- Rhs of binding (or arg of lambda)
1067 -> ([InBndr], InExpr) -- Body of the let/lambda
1070 -> SimplM (SimplEnv, OutExpr)
1072 -- simplNonRecE is used for
1073 -- * non-top-level non-recursive lets in expressions
1076 -- It deals with strict bindings, via the StrictBind continuation,
1077 -- which may abort the whole process
1079 -- The "body" of the binding comes as a pair of ([InId],InExpr)
1080 -- representing a lambda; so we recurse back to simplLam
1081 -- Why? Because of the binder-occ-info-zapping done before
1082 -- the call to simplLam in simplExprF (Lam ...)
1084 -- First deal with type applications and type lets
1085 -- (/\a. e) (Type ty) and (let a = Type ty in e)
1086 simplNonRecE env bndr (Type ty_arg, rhs_se) (bndrs, body) cont
1087 = ASSERT( isTyCoVar bndr )
1088 do { ty_arg' <- simplType (rhs_se `setInScope` env) ty_arg
1089 ; simplLam (extendTvSubst env bndr ty_arg') bndrs body cont }
1091 simplNonRecE env bndr (rhs, rhs_se) (bndrs, body) cont
1092 | preInlineUnconditionally env NotTopLevel bndr rhs
1093 = do { tick (PreInlineUnconditionally bndr)
1094 ; -- pprTrace "preInlineUncond" (ppr bndr <+> ppr rhs) $
1095 simplLam (extendIdSubst env bndr (mkContEx rhs_se rhs)) bndrs body cont }
1098 = do { simplExprF (rhs_se `setFloats` env) rhs
1099 (StrictBind bndr bndrs body env cont) }
1102 = ASSERT( not (isTyCoVar bndr) )
1103 do { (env1, bndr1) <- simplNonRecBndr env bndr
1104 ; let (env2, bndr2) = addBndrRules env1 bndr bndr1
1105 ; env3 <- simplLazyBind env2 NotTopLevel NonRecursive bndr bndr2 rhs rhs_se
1106 ; simplLam env3 bndrs body cont }
1110 %************************************************************************
1114 %************************************************************************
1117 -- Hack alert: we only distinguish subsumed cost centre stacks for the
1118 -- purposes of inlining. All other CCCSs are mapped to currentCCS.
1119 simplNote :: SimplEnv -> Note -> CoreExpr -> SimplCont
1120 -> SimplM (SimplEnv, OutExpr)
1121 simplNote env (SCC cc) e cont
1122 | pushCCisNop cc (getEnclosingCC env) -- scc "f" (...(scc "f" e)...)
1123 = simplExprF env e cont -- ==> scc "f" (...e...)
1125 = do { e' <- simplExpr (setEnclosingCC env currentCCS) e
1126 ; rebuild env (mkSCC cc e') cont }
1128 simplNote env (CoreNote s) e cont
1129 = do { e' <- simplExpr env e
1130 ; rebuild env (Note (CoreNote s) e') cont }
1134 %************************************************************************
1138 %************************************************************************
1141 simplVar :: SimplEnv -> InVar -> SimplM OutExpr
1142 -- Look up an InVar in the environment
1145 = return (Type (substTyVar env var))
1147 = case substId env var of
1148 DoneId var1 -> return (Var var1)
1149 DoneEx e -> return e
1150 ContEx tvs ids e -> simplExpr (setSubstEnv env tvs ids) e
1152 simplVarF :: SimplEnv -> InId -> SimplCont -> SimplM (SimplEnv, OutExpr)
1153 simplVarF env var cont
1154 = case substId env var of
1155 DoneEx e -> simplExprF (zapSubstEnv env) e cont
1156 ContEx tvs ids e -> simplExprF (setSubstEnv env tvs ids) e cont
1157 DoneId var1 -> completeCall env var1 cont
1158 -- Note [zapSubstEnv]
1159 -- The template is already simplified, so don't re-substitute.
1160 -- This is VITAL. Consider
1162 -- let y = \z -> ...x... in
1164 -- We'll clone the inner \x, adding x->x' in the id_subst
1165 -- Then when we inline y, we must *not* replace x by x' in
1166 -- the inlined copy!!
1168 ---------------------------------------------------------
1169 -- Dealing with a call site
1171 completeCall :: SimplEnv -> Id -> SimplCont -> SimplM (SimplEnv, OutExpr)
1172 completeCall env var cont
1173 = do { ------------- Try inlining ----------------
1174 dflags <- getDOptsSmpl
1175 ; let (lone_variable, arg_infos, call_cont) = contArgs cont
1176 -- The args are OutExprs, obtained by *lazily* substituting
1177 -- in the args found in cont. These args are only examined
1178 -- to limited depth (unless a rule fires). But we must do
1179 -- the substitution; rule matching on un-simplified args would
1182 n_val_args = length arg_infos
1183 interesting_cont = interestingCallContext call_cont
1184 unfolding = activeUnfolding env var
1185 maybe_inline = callSiteInline dflags var unfolding
1186 lone_variable arg_infos interesting_cont
1187 ; case maybe_inline of {
1188 Just expr -- There is an inlining!
1189 -> do { tick (UnfoldingDone var)
1190 ; trace_inline dflags expr cont $
1191 simplExprF (zapSubstEnv env) expr cont }
1193 ; Nothing -> do -- No inlining!
1195 { rule_base <- getSimplRules
1196 ; let info = mkArgInfo var (getRules rule_base var) n_val_args call_cont
1197 ; rebuildCall env info cont
1200 trace_inline dflags unfolding cont stuff
1201 | not (dopt Opt_D_dump_inlinings dflags) = stuff
1202 | not (dopt Opt_D_verbose_core2core dflags)
1203 = if isExternalName (idName var) then
1204 pprTrace "Inlining done:" (ppr var) stuff
1207 = pprTrace ("Inlining done: " ++ showSDoc (ppr var))
1208 (vcat [text "Inlined fn: " <+> nest 2 (ppr unfolding),
1209 text "Cont: " <+> ppr cont])
1212 rebuildCall :: SimplEnv
1215 -> SimplM (SimplEnv, OutExpr)
1216 rebuildCall env (ArgInfo { ai_fun = fun, ai_args = rev_args, ai_strs = [] }) cont
1217 -- When we run out of strictness args, it means
1218 -- that the call is definitely bottom; see SimplUtils.mkArgInfo
1219 -- Then we want to discard the entire strict continuation. E.g.
1220 -- * case (error "hello") of { ... }
1221 -- * (error "Hello") arg
1222 -- * f (error "Hello") where f is strict
1224 -- Then, especially in the first of these cases, we'd like to discard
1225 -- the continuation, leaving just the bottoming expression. But the
1226 -- type might not be right, so we may have to add a coerce.
1227 | not (contIsTrivial cont) -- Only do this if there is a non-trivial
1228 = return (env, mk_coerce res) -- contination to discard, else we do it
1229 where -- again and again!
1230 res = mkApps (Var fun) (reverse rev_args)
1231 res_ty = exprType res
1232 cont_ty = contResultType env res_ty cont
1233 co = mkUnsafeCoercion res_ty cont_ty
1234 mk_coerce expr | cont_ty `coreEqType` res_ty = expr
1235 | otherwise = mkCoerce co expr
1237 rebuildCall env info (ApplyTo _ (Type arg_ty) se cont)
1238 = do { ty' <- simplCoercion (se `setInScope` env) arg_ty
1239 ; rebuildCall env (info `addArgTo` Type ty') cont }
1241 rebuildCall env info@(ArgInfo { ai_encl = encl_rules
1242 , ai_strs = str:strs, ai_discs = disc:discs })
1243 (ApplyTo dup_flag arg arg_se cont)
1244 | isSimplified dup_flag -- See Note [Avoid redundant simplification]
1245 = rebuildCall env (addArgTo info' arg) cont
1247 | str -- Strict argument
1248 = -- pprTrace "Strict Arg" (ppr arg $$ ppr (seIdSubst env) $$ ppr (seInScope env)) $
1249 simplExprF (arg_se `setFloats` env) arg
1250 (StrictArg info' cci cont)
1253 | otherwise -- Lazy argument
1254 -- DO NOT float anything outside, hence simplExprC
1255 -- There is no benefit (unlike in a let-binding), and we'd
1256 -- have to be very careful about bogus strictness through
1257 -- floating a demanded let.
1258 = do { arg' <- simplExprC (arg_se `setInScope` env) arg
1260 ; rebuildCall env (addArgTo info' arg') cont }
1262 info' = info { ai_strs = strs, ai_discs = discs }
1263 cci | encl_rules || disc > 0 = ArgCtxt encl_rules -- Be keener here
1264 | otherwise = BoringCtxt -- Nothing interesting
1266 rebuildCall env (ArgInfo { ai_fun = fun, ai_args = rev_args, ai_rules = rules }) cont
1267 = do { -- We've accumulated a simplified call in <fun,rev_args>
1268 -- so try rewrite rules; see Note [RULEs apply to simplified arguments]
1269 -- See also Note [Rules for recursive functions]
1270 ; let args = reverse rev_args
1271 env' = zapSubstEnv env
1272 ; mb_rule <- tryRules env rules fun args cont
1274 Just (n_args, rule_rhs) -> simplExprF env' rule_rhs $
1275 pushSimplifiedArgs env' (drop n_args args) cont ;
1276 -- n_args says how many args the rule consumed
1277 ; Nothing -> rebuild env (mkApps (Var fun) args) cont -- No rules
1281 Note [RULES apply to simplified arguments]
1282 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1283 It's very desirable to try RULES once the arguments have been simplified, because
1284 doing so ensures that rule cascades work in one pass. Consider
1285 {-# RULES g (h x) = k x
1288 Then we want to rewrite (g (h x)) to (k x) and only then try f's rules. If
1289 we match f's rules against the un-simplified RHS, it won't match. This
1290 makes a particularly big difference when superclass selectors are involved:
1291 op ($p1 ($p2 (df d)))
1292 We want all this to unravel in one sweeep.
1294 Note [Avoid redundant simplification]
1295 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1296 Because RULES apply to simplified arguments, there's a danger of repeatedly
1297 simplifying already-simplified arguments. An important example is that of
1299 Here e1, e2 are simplified before the rule is applied, but don't really
1300 participate in the rule firing. So we mark them as Simplified to avoid
1301 re-simplifying them.
1305 This part of the simplifier may break the no-shadowing invariant
1307 f (...(\a -> e)...) (case y of (a,b) -> e')
1308 where f is strict in its second arg
1309 If we simplify the innermost one first we get (...(\a -> e)...)
1310 Simplifying the second arg makes us float the case out, so we end up with
1311 case y of (a,b) -> f (...(\a -> e)...) e'
1312 So the output does not have the no-shadowing invariant. However, there is
1313 no danger of getting name-capture, because when the first arg was simplified
1314 we used an in-scope set that at least mentioned all the variables free in its
1315 static environment, and that is enough.
1317 We can't just do innermost first, or we'd end up with a dual problem:
1318 case x of (a,b) -> f e (...(\a -> e')...)
1320 I spent hours trying to recover the no-shadowing invariant, but I just could
1321 not think of an elegant way to do it. The simplifier is already knee-deep in
1322 continuations. We have to keep the right in-scope set around; AND we have
1323 to get the effect that finding (error "foo") in a strict arg position will
1324 discard the entire application and replace it with (error "foo"). Getting
1325 all this at once is TOO HARD!
1328 %************************************************************************
1332 %************************************************************************
1335 tryRules :: SimplEnv -> [CoreRule]
1336 -> Id -> [OutExpr] -> SimplCont
1337 -> SimplM (Maybe (Arity, CoreExpr)) -- The arity is the number of
1338 -- args consumed by the rule
1339 tryRules env rules fn args call_cont
1343 = do { dflags <- getDOptsSmpl
1344 ; case activeRule dflags env of {
1345 Nothing -> return Nothing ; -- No rules apply
1347 case lookupRule act_fn (activeUnfInRule env) (getInScope env) fn args rules of {
1348 Nothing -> return Nothing ; -- No rule matches
1349 Just (rule, rule_rhs) ->
1351 do { tick (RuleFired (ru_name rule))
1352 ; trace_dump dflags rule rule_rhs $
1353 return (Just (ruleArity rule, rule_rhs)) }}}}
1355 trace_dump dflags rule rule_rhs stuff
1356 | not (dopt Opt_D_dump_rule_firings dflags) = stuff
1357 | not (dopt Opt_D_verbose_core2core dflags)
1359 = pprTrace "Rule fired:" (ftext (ru_name rule)) stuff
1361 = pprTrace "Rule fired"
1362 (vcat [text "Rule:" <+> ftext (ru_name rule),
1363 text "Before:" <+> ppr fn <+> sep (map pprParendExpr args),
1364 text "After: " <+> pprCoreExpr rule_rhs,
1365 text "Cont: " <+> ppr call_cont])
1369 Note [Rules for recursive functions]
1370 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1371 You might think that we shouldn't apply rules for a loop breaker:
1372 doing so might give rise to an infinite loop, because a RULE is
1373 rather like an extra equation for the function:
1374 RULE: f (g x) y = x+y
1377 But it's too drastic to disable rules for loop breakers.
1378 Even the foldr/build rule would be disabled, because foldr
1379 is recursive, and hence a loop breaker:
1380 foldr k z (build g) = g k z
1381 So it's up to the programmer: rules can cause divergence
1384 %************************************************************************
1386 Rebuilding a cse expression
1388 %************************************************************************
1390 Note [Case elimination]
1391 ~~~~~~~~~~~~~~~~~~~~~~~
1392 The case-elimination transformation discards redundant case expressions.
1393 Start with a simple situation:
1395 case x# of ===> e[x#/y#]
1398 (when x#, y# are of primitive type, of course). We can't (in general)
1399 do this for algebraic cases, because we might turn bottom into
1402 The code in SimplUtils.prepareAlts has the effect of generalise this
1403 idea to look for a case where we're scrutinising a variable, and we
1404 know that only the default case can match. For example:
1408 DEFAULT -> ...(case x of
1412 Here the inner case is first trimmed to have only one alternative, the
1413 DEFAULT, after which it's an instance of the previous case. This
1414 really only shows up in eliminating error-checking code.
1416 We also make sure that we deal with this very common case:
1421 Here we are using the case as a strict let; if x is used only once
1422 then we want to inline it. We have to be careful that this doesn't
1423 make the program terminate when it would have diverged before, so we
1425 - e is already evaluated (it may so if e is a variable)
1426 - x is used strictly, or
1428 Lastly, the code in SimplUtils.mkCase combines identical RHSs. So
1430 case e of ===> case e of DEFAULT -> r
1434 Now again the case may be elminated by the CaseElim transformation.
1437 Further notes about case elimination
1438 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1439 Consider: test :: Integer -> IO ()
1442 Turns out that this compiles to:
1445 eta1 :: State# RealWorld ->
1446 case PrelNum.< eta PrelNum.zeroInteger of wild { __DEFAULT ->
1448 (PrelNum.jtos eta ($w[] @ Char))
1450 of wild1 { (# new_s, a4 #) -> PrelIO.lvl23 new_s }}
1452 Notice the strange '<' which has no effect at all. This is a funny one.
1453 It started like this:
1455 f x y = if x < 0 then jtos x
1456 else if y==0 then "" else jtos x
1458 At a particular call site we have (f v 1). So we inline to get
1460 if v < 0 then jtos x
1461 else if 1==0 then "" else jtos x
1463 Now simplify the 1==0 conditional:
1465 if v<0 then jtos v else jtos v
1467 Now common-up the two branches of the case:
1469 case (v<0) of DEFAULT -> jtos v
1471 Why don't we drop the case? Because it's strict in v. It's technically
1472 wrong to drop even unnecessary evaluations, and in practice they
1473 may be a result of 'seq' so we *definitely* don't want to drop those.
1474 I don't really know how to improve this situation.
1477 ---------------------------------------------------------
1478 -- Eliminate the case if possible
1480 rebuildCase, reallyRebuildCase
1482 -> OutExpr -- Scrutinee
1483 -> InId -- Case binder
1484 -> [InAlt] -- Alternatives (inceasing order)
1486 -> SimplM (SimplEnv, OutExpr)
1488 --------------------------------------------------
1489 -- 1. Eliminate the case if there's a known constructor
1490 --------------------------------------------------
1492 rebuildCase env scrut case_bndr alts cont
1493 | Lit lit <- scrut -- No need for same treatment as constructors
1494 -- because literals are inlined more vigorously
1495 = do { tick (KnownBranch case_bndr)
1496 ; case findAlt (LitAlt lit) alts of
1497 Nothing -> missingAlt env case_bndr alts cont
1498 Just (_, bs, rhs) -> simple_rhs bs rhs }
1500 | Just (con, ty_args, other_args) <- exprIsConApp_maybe (activeUnfInRule env) scrut
1501 -- Works when the scrutinee is a variable with a known unfolding
1502 -- as well as when it's an explicit constructor application
1503 = do { tick (KnownBranch case_bndr)
1504 ; case findAlt (DataAlt con) alts of
1505 Nothing -> missingAlt env case_bndr alts cont
1506 Just (DEFAULT, bs, rhs) -> simple_rhs bs rhs
1507 Just (_, bs, rhs) -> knownCon env scrut con ty_args other_args
1508 case_bndr bs rhs cont
1511 simple_rhs bs rhs = ASSERT( null bs )
1512 do { env' <- simplNonRecX env case_bndr scrut
1513 ; simplExprF env' rhs cont }
1516 --------------------------------------------------
1517 -- 2. Eliminate the case if scrutinee is evaluated
1518 --------------------------------------------------
1520 rebuildCase env scrut case_bndr [(_, bndrs, rhs)] cont
1521 -- See if we can get rid of the case altogether
1522 -- See Note [Case elimination]
1523 -- mkCase made sure that if all the alternatives are equal,
1524 -- then there is now only one (DEFAULT) rhs
1525 | all isDeadBinder bndrs -- bndrs are [InId]
1527 -- Check that the scrutinee can be let-bound instead of case-bound
1528 , exprOkForSpeculation scrut
1529 -- OK not to evaluate it
1530 -- This includes things like (==# a# b#)::Bool
1531 -- so that we simplify
1532 -- case ==# a# b# of { True -> x; False -> x }
1535 -- This particular example shows up in default methods for
1536 -- comparision operations (e.g. in (>=) for Int.Int32)
1537 || exprIsHNF scrut -- It's already evaluated
1538 || var_demanded_later scrut -- It'll be demanded later
1540 -- || not opt_SimplPedanticBottoms) -- Or we don't care!
1541 -- We used to allow improving termination by discarding cases, unless -fpedantic-bottoms was on,
1542 -- but that breaks badly for the dataToTag# primop, which relies on a case to evaluate
1543 -- its argument: case x of { y -> dataToTag# y }
1544 -- Here we must *not* discard the case, because dataToTag# just fetches the tag from
1545 -- the info pointer. So we'll be pedantic all the time, and see if that gives any
1547 -- Also we don't want to discard 'seq's
1548 = do { tick (CaseElim case_bndr)
1549 ; env' <- simplNonRecX env case_bndr scrut
1550 ; simplExprF env' rhs cont }
1552 -- The case binder is going to be evaluated later,
1553 -- and the scrutinee is a simple variable
1554 var_demanded_later (Var v) = isStrictDmd (idDemandInfo case_bndr)
1555 && not (isTickBoxOp v)
1556 -- ugly hack; covering this case is what
1557 -- exprOkForSpeculation was intended for.
1558 var_demanded_later _ = False
1560 --------------------------------------------------
1561 -- 3. Try seq rules; see Note [User-defined RULES for seq] in MkId
1562 --------------------------------------------------
1564 rebuildCase env scrut case_bndr alts@[(_, bndrs, rhs)] cont
1565 | all isDeadBinder (case_bndr : bndrs) -- So this is just 'seq'
1566 = do { let rhs' = substExpr (text "rebuild-case") env rhs
1567 out_args = [Type (substTy env (idType case_bndr)),
1568 Type (exprType rhs'), scrut, rhs']
1569 -- Lazily evaluated, so we don't do most of this
1571 ; rule_base <- getSimplRules
1572 ; mb_rule <- tryRules env (getRules rule_base seqId) seqId out_args cont
1574 Just (n_args, res) -> simplExprF (zapSubstEnv env)
1575 (mkApps res (drop n_args out_args))
1577 Nothing -> reallyRebuildCase env scrut case_bndr alts cont }
1579 rebuildCase env scrut case_bndr alts cont
1580 = reallyRebuildCase env scrut case_bndr alts cont
1582 --------------------------------------------------
1583 -- 3. Catch-all case
1584 --------------------------------------------------
1586 reallyRebuildCase env scrut case_bndr alts cont
1587 = do { -- Prepare the continuation;
1588 -- The new subst_env is in place
1589 (env', dup_cont, nodup_cont) <- prepareCaseCont env alts cont
1591 -- Simplify the alternatives
1592 ; (scrut', case_bndr', alts') <- simplAlts env' scrut case_bndr alts dup_cont
1594 -- Check for empty alternatives
1595 ; if null alts' then missingAlt env case_bndr alts cont
1597 { dflags <- getDOptsSmpl
1598 ; case_expr <- mkCase dflags scrut' case_bndr' alts'
1600 -- Notice that rebuild gets the in-scope set from env', not alt_env
1601 -- (which in any case is only build in simplAlts)
1602 -- The case binder *not* scope over the whole returned case-expression
1603 ; rebuild env' case_expr nodup_cont } }
1606 simplCaseBinder checks whether the scrutinee is a variable, v. If so,
1607 try to eliminate uses of v in the RHSs in favour of case_bndr; that
1608 way, there's a chance that v will now only be used once, and hence
1611 Historical note: we use to do the "case binder swap" in the Simplifier
1612 so there were additional complications if the scrutinee was a variable.
1613 Now the binder-swap stuff is done in the occurrence analyer; see
1614 OccurAnal Note [Binder swap].
1618 If the case binder is not dead, then neither are the pattern bound
1620 case <any> of x { (a,b) ->
1621 case x of { (p,q) -> p } }
1622 Here (a,b) both look dead, but come alive after the inner case is eliminated.
1623 The point is that we bring into the envt a binding
1625 after the outer case, and that makes (a,b) alive. At least we do unless
1626 the case binder is guaranteed dead.
1628 In practice, the scrutinee is almost always a variable, so we pretty
1629 much always zap the OccInfo of the binders. It doesn't matter much though.
1634 Consider case (v `cast` co) of x { I# y ->
1635 ... (case (v `cast` co) of {...}) ...
1636 We'd like to eliminate the inner case. We can get this neatly by
1637 arranging that inside the outer case we add the unfolding
1638 v |-> x `cast` (sym co)
1639 to v. Then we should inline v at the inner case, cancel the casts, and away we go
1641 Note [Improving seq]
1644 type family F :: * -> *
1645 type instance F Int = Int
1647 ... case e of x { DEFAULT -> rhs } ...
1649 where x::F Int. Then we'd like to rewrite (F Int) to Int, getting
1651 case e `cast` co of x'::Int
1652 I# x# -> let x = x' `cast` sym co
1655 so that 'rhs' can take advantage of the form of x'.
1657 Notice that Note [Case of cast] may then apply to the result.
1659 Nota Bene: We only do the [Improving seq] transformation if the
1660 case binder 'x' is actually used in the rhs; that is, if the case
1661 is *not* a *pure* seq.
1662 a) There is no point in adding the cast to a pure seq.
1663 b) There is a good reason not to: doing so would interfere
1664 with seq rules (Note [Built-in RULES for seq] in MkId).
1665 In particular, this [Improving seq] thing *adds* a cast
1666 while [Built-in RULES for seq] *removes* one, so they
1669 You might worry about
1670 case v of x { __DEFAULT ->
1671 ... case (v `cast` co) of y { I# -> ... }}
1672 This is a pure seq (since x is unused), so [Improving seq] won't happen.
1673 But it's ok: the simplifier will replace 'v' by 'x' in the rhs to get
1674 case v of x { __DEFAULT ->
1675 ... case (x `cast` co) of y { I# -> ... }}
1676 Now the outer case is not a pure seq, so [Improving seq] will happen,
1677 and then the inner case will disappear.
1679 The need for [Improving seq] showed up in Roman's experiments. Example:
1680 foo :: F Int -> Int -> Int
1681 foo t n = t `seq` bar n
1684 bar n = bar (n - case t of TI i -> i)
1685 Here we'd like to avoid repeated evaluating t inside the loop, by
1686 taking advantage of the `seq`.
1688 At one point I did transformation in LiberateCase, but it's more
1689 robust here. (Otherwise, there's a danger that we'll simply drop the
1690 'seq' altogether, before LiberateCase gets to see it.)
1693 simplAlts :: SimplEnv
1695 -> InId -- Case binder
1696 -> [InAlt] -- Non-empty
1698 -> SimplM (OutExpr, OutId, [OutAlt]) -- Includes the continuation
1699 -- Like simplExpr, this just returns the simplified alternatives;
1700 -- it does not return an environment
1702 simplAlts env scrut case_bndr alts cont'
1703 = -- pprTrace "simplAlts" (ppr alts $$ ppr (seTvSubst env)) $
1704 do { let env0 = zapFloats env
1706 ; (env1, case_bndr1) <- simplBinder env0 case_bndr
1708 ; fam_envs <- getFamEnvs
1709 ; (alt_env', scrut', case_bndr') <- improveSeq fam_envs env1 scrut
1710 case_bndr case_bndr1 alts
1712 ; (imposs_deflt_cons, in_alts) <- prepareAlts scrut' case_bndr' alts
1714 ; alts' <- mapM (simplAlt alt_env' imposs_deflt_cons case_bndr' cont') in_alts
1715 ; return (scrut', case_bndr', alts') }
1718 ------------------------------------
1719 improveSeq :: (FamInstEnv, FamInstEnv) -> SimplEnv
1720 -> OutExpr -> InId -> OutId -> [InAlt]
1721 -> SimplM (SimplEnv, OutExpr, OutId)
1722 -- Note [Improving seq]
1723 improveSeq fam_envs env scrut case_bndr case_bndr1 [(DEFAULT,_,_)]
1724 | not (isDeadBinder case_bndr) -- Not a pure seq! See the Note!
1725 , Just (co, ty2) <- topNormaliseType fam_envs (idType case_bndr1)
1726 = do { case_bndr2 <- newId (fsLit "nt") ty2
1727 ; let rhs = DoneEx (Var case_bndr2 `Cast` mkSymCoercion co)
1728 env2 = extendIdSubst env case_bndr rhs
1729 ; return (env2, scrut `Cast` co, case_bndr2) }
1731 improveSeq _ env scrut _ case_bndr1 _
1732 = return (env, scrut, case_bndr1)
1735 ------------------------------------
1736 simplAlt :: SimplEnv
1737 -> [AltCon] -- These constructors can't be present when
1738 -- matching the DEFAULT alternative
1739 -> OutId -- The case binder
1744 simplAlt env imposs_deflt_cons case_bndr' cont' (DEFAULT, bndrs, rhs)
1745 = ASSERT( null bndrs )
1746 do { let env' = addBinderOtherCon env case_bndr' imposs_deflt_cons
1747 -- Record the constructors that the case-binder *can't* be.
1748 ; rhs' <- simplExprC env' rhs cont'
1749 ; return (DEFAULT, [], rhs') }
1751 simplAlt env _ case_bndr' cont' (LitAlt lit, bndrs, rhs)
1752 = ASSERT( null bndrs )
1753 do { let env' = addBinderUnfolding env case_bndr' (Lit lit)
1754 ; rhs' <- simplExprC env' rhs cont'
1755 ; return (LitAlt lit, [], rhs') }
1757 simplAlt env _ case_bndr' cont' (DataAlt con, vs, rhs)
1758 = do { -- Deal with the pattern-bound variables
1759 -- Mark the ones that are in ! positions in the
1760 -- data constructor as certainly-evaluated.
1761 -- NB: simplLamBinders preserves this eval info
1762 let vs_with_evals = add_evals (dataConRepStrictness con)
1763 ; (env', vs') <- simplLamBndrs env vs_with_evals
1765 -- Bind the case-binder to (con args)
1766 ; let inst_tys' = tyConAppArgs (idType case_bndr')
1767 con_args = map Type inst_tys' ++ varsToCoreExprs vs'
1768 env'' = addBinderUnfolding env' case_bndr'
1769 (mkConApp con con_args)
1771 ; rhs' <- simplExprC env'' rhs cont'
1772 ; return (DataAlt con, vs', rhs') }
1774 -- add_evals records the evaluated-ness of the bound variables of
1775 -- a case pattern. This is *important*. Consider
1776 -- data T = T !Int !Int
1778 -- case x of { T a b -> T (a+1) b }
1780 -- We really must record that b is already evaluated so that we don't
1781 -- go and re-evaluate it when constructing the result.
1782 -- See Note [Data-con worker strictness] in MkId.lhs
1787 go (v:vs') strs | isTyCoVar v = v : go vs' strs
1788 go (v:vs') (str:strs)
1789 | isMarkedStrict str = evald_v : go vs' strs
1790 | otherwise = zapped_v : go vs' strs
1792 zapped_v = zap_occ_info v
1793 evald_v = zapped_v `setIdUnfolding` evaldUnfolding
1794 go _ _ = pprPanic "cat_evals" (ppr con $$ ppr vs $$ ppr the_strs)
1796 -- See Note [zapOccInfo]
1797 -- zap_occ_info: if the case binder is alive, then we add the unfolding
1799 -- to the envt; so vs are now very much alive
1800 -- Note [Aug06] I can't see why this actually matters, but it's neater
1801 -- case e of t { (a,b) -> ...(case t of (p,q) -> p)... }
1802 -- ==> case e of t { (a,b) -> ...(a)... }
1803 -- Look, Ma, a is alive now.
1804 zap_occ_info = zapCasePatIdOcc case_bndr'
1806 addBinderUnfolding :: SimplEnv -> Id -> CoreExpr -> SimplEnv
1807 addBinderUnfolding env bndr rhs
1808 = modifyInScope env (bndr `setIdUnfolding` mkSimpleUnfolding rhs)
1810 addBinderOtherCon :: SimplEnv -> Id -> [AltCon] -> SimplEnv
1811 addBinderOtherCon env bndr cons
1812 = modifyInScope env (bndr `setIdUnfolding` mkOtherCon cons)
1814 zapCasePatIdOcc :: Id -> Id -> Id
1815 -- Consider case e of b { (a,b) -> ... }
1816 -- Then if we bind b to (a,b) in "...", and b is not dead,
1817 -- then we must zap the deadness info on a,b
1818 zapCasePatIdOcc case_bndr
1819 | isDeadBinder case_bndr = \ pat_id -> pat_id
1820 | otherwise = \ pat_id -> zapIdOccInfo pat_id
1824 %************************************************************************
1826 \subsection{Known constructor}
1828 %************************************************************************
1830 We are a bit careful with occurrence info. Here's an example
1832 (\x* -> case x of (a*, b) -> f a) (h v, e)
1834 where the * means "occurs once". This effectively becomes
1835 case (h v, e) of (a*, b) -> f a)
1837 let a* = h v; b = e in f a
1841 All this should happen in one sweep.
1844 knownCon :: SimplEnv
1845 -> OutExpr -- The scrutinee
1846 -> DataCon -> [OutType] -> [OutExpr] -- The scrutinee (in pieces)
1847 -> InId -> [InBndr] -> InExpr -- The alternative
1849 -> SimplM (SimplEnv, OutExpr)
1851 knownCon env scrut dc dc_ty_args dc_args bndr bs rhs cont
1852 = do { env' <- bind_args env bs dc_args
1853 ; env'' <- bind_case_bndr env'
1854 ; simplExprF env'' rhs cont }
1856 zap_occ = zapCasePatIdOcc bndr -- bndr is an InId
1859 bind_args env' [] _ = return env'
1861 bind_args env' (b:bs') (Type ty : args)
1862 = ASSERT( isTyCoVar b )
1863 bind_args (extendTvSubst env' b ty) bs' args
1865 bind_args env' (b:bs') (arg : args)
1867 do { let b' = zap_occ b
1868 -- Note that the binder might be "dead", because it doesn't
1869 -- occur in the RHS; and simplNonRecX may therefore discard
1870 -- it via postInlineUnconditionally.
1871 -- Nevertheless we must keep it if the case-binder is alive,
1872 -- because it may be used in the con_app. See Note [zapOccInfo]
1873 ; env'' <- simplNonRecX env' b' arg
1874 ; bind_args env'' bs' args }
1877 pprPanic "bind_args" $ ppr dc $$ ppr bs $$ ppr dc_args $$
1878 text "scrut:" <+> ppr scrut
1880 -- It's useful to bind bndr to scrut, rather than to a fresh
1881 -- binding x = Con arg1 .. argn
1882 -- because very often the scrut is a variable, so we avoid
1883 -- creating, and then subsequently eliminating, a let-binding
1884 -- BUT, if scrut is a not a variable, we must be careful
1885 -- about duplicating the arg redexes; in that case, make
1886 -- a new con-app from the args
1888 | isDeadBinder bndr = return env
1889 | exprIsTrivial scrut = return (extendIdSubst env bndr (DoneEx scrut))
1890 | otherwise = do { dc_args <- mapM (simplVar env) bs
1891 -- dc_ty_args are aready OutTypes,
1892 -- but bs are InBndrs
1893 ; let con_app = Var (dataConWorkId dc)
1894 `mkTyApps` dc_ty_args
1896 ; simplNonRecX env bndr con_app }
1899 missingAlt :: SimplEnv -> Id -> [InAlt] -> SimplCont -> SimplM (SimplEnv, OutExpr)
1900 -- This isn't strictly an error, although it is unusual.
1901 -- It's possible that the simplifer might "see" that
1902 -- an inner case has no accessible alternatives before
1903 -- it "sees" that the entire branch of an outer case is
1904 -- inaccessible. So we simply put an error case here instead.
1905 missingAlt env case_bndr alts cont
1906 = WARN( True, ptext (sLit "missingAlt") <+> ppr case_bndr )
1907 return (env, mkImpossibleExpr res_ty)
1909 res_ty = contResultType env (substTy env (coreAltsType alts)) cont
1913 %************************************************************************
1915 \subsection{Duplicating continuations}
1917 %************************************************************************
1920 prepareCaseCont :: SimplEnv
1921 -> [InAlt] -> SimplCont
1922 -> SimplM (SimplEnv, SimplCont,SimplCont)
1923 -- Return a duplicatable continuation, a non-duplicable part
1924 -- plus some extra bindings (that scope over the entire
1927 -- No need to make it duplicatable if there's only one alternative
1928 prepareCaseCont env [_] cont = return (env, cont, mkBoringStop)
1929 prepareCaseCont env _ cont = mkDupableCont env cont
1933 mkDupableCont :: SimplEnv -> SimplCont
1934 -> SimplM (SimplEnv, SimplCont, SimplCont)
1936 mkDupableCont env cont
1937 | contIsDupable cont
1938 = return (env, cont, mkBoringStop)
1940 mkDupableCont _ (Stop {}) = panic "mkDupableCont" -- Handled by previous eqn
1942 mkDupableCont env (CoerceIt ty cont)
1943 = do { (env', dup, nodup) <- mkDupableCont env cont
1944 ; return (env', CoerceIt ty dup, nodup) }
1946 mkDupableCont env cont@(StrictBind {})
1947 = return (env, mkBoringStop, cont)
1948 -- See Note [Duplicating StrictBind]
1950 mkDupableCont env (StrictArg info cci cont)
1951 -- See Note [Duplicating StrictArg]
1952 = do { (env', dup, nodup) <- mkDupableCont env cont
1953 ; (env'', args') <- mapAccumLM (makeTrivial NotTopLevel) env' (ai_args info)
1954 ; return (env'', StrictArg (info { ai_args = args' }) cci dup, nodup) }
1956 mkDupableCont env (ApplyTo _ arg se cont)
1957 = -- e.g. [...hole...] (...arg...)
1959 -- let a = ...arg...
1960 -- in [...hole...] a
1961 do { (env', dup_cont, nodup_cont) <- mkDupableCont env cont
1962 ; arg' <- simplExpr (se `setInScope` env') arg
1963 ; (env'', arg'') <- makeTrivial NotTopLevel env' arg'
1964 ; let app_cont = ApplyTo OkToDup arg'' (zapSubstEnv env'') dup_cont
1965 ; return (env'', app_cont, nodup_cont) }
1967 mkDupableCont env cont@(Select _ case_bndr [(_, bs, _rhs)] _ _)
1968 -- See Note [Single-alternative case]
1969 -- | not (exprIsDupable rhs && contIsDupable case_cont)
1970 -- | not (isDeadBinder case_bndr)
1971 | all isDeadBinder bs -- InIds
1972 && not (isUnLiftedType (idType case_bndr))
1973 -- Note [Single-alternative-unlifted]
1974 = return (env, mkBoringStop, cont)
1976 mkDupableCont env (Select _ case_bndr alts se cont)
1977 = -- e.g. (case [...hole...] of { pi -> ei })
1979 -- let ji = \xij -> ei
1980 -- in case [...hole...] of { pi -> ji xij }
1981 do { tick (CaseOfCase case_bndr)
1982 ; (env', dup_cont, nodup_cont) <- mkDupableCont env cont
1983 -- NB: call mkDupableCont here, *not* prepareCaseCont
1984 -- We must make a duplicable continuation, whereas prepareCaseCont
1985 -- doesn't when there is a single case branch
1987 ; let alt_env = se `setInScope` env'
1988 ; (alt_env', case_bndr') <- simplBinder alt_env case_bndr
1989 ; alts' <- mapM (simplAlt alt_env' [] case_bndr' dup_cont) alts
1990 -- Safe to say that there are no handled-cons for the DEFAULT case
1991 -- NB: simplBinder does not zap deadness occ-info, so
1992 -- a dead case_bndr' will still advertise its deadness
1993 -- This is really important because in
1994 -- case e of b { (# p,q #) -> ... }
1995 -- b is always dead, and indeed we are not allowed to bind b to (# p,q #),
1996 -- which might happen if e was an explicit unboxed pair and b wasn't marked dead.
1997 -- In the new alts we build, we have the new case binder, so it must retain
1999 -- NB: we don't use alt_env further; it has the substEnv for
2000 -- the alternatives, and we don't want that
2002 ; (env'', alts'') <- mkDupableAlts env' case_bndr' alts'
2003 ; return (env'', -- Note [Duplicated env]
2004 Select OkToDup case_bndr' alts'' (zapSubstEnv env'') mkBoringStop,
2008 mkDupableAlts :: SimplEnv -> OutId -> [InAlt]
2009 -> SimplM (SimplEnv, [InAlt])
2010 -- Absorbs the continuation into the new alternatives
2012 mkDupableAlts env case_bndr' the_alts
2015 go env0 [] = return (env0, [])
2017 = do { (env1, alt') <- mkDupableAlt env0 case_bndr' alt
2018 ; (env2, alts') <- go env1 alts
2019 ; return (env2, alt' : alts' ) }
2021 mkDupableAlt :: SimplEnv -> OutId -> (AltCon, [CoreBndr], CoreExpr)
2022 -> SimplM (SimplEnv, (AltCon, [CoreBndr], CoreExpr))
2023 mkDupableAlt env case_bndr (con, bndrs', rhs')
2024 | exprIsDupable rhs' -- Note [Small alternative rhs]
2025 = return (env, (con, bndrs', rhs'))
2027 = do { let rhs_ty' = exprType rhs'
2028 scrut_ty = idType case_bndr
2031 DEFAULT -> case_bndr
2032 DataAlt dc -> setIdUnfolding case_bndr unf
2034 -- See Note [Case binders and join points]
2035 unf = mkInlineUnfolding Nothing rhs
2036 rhs = mkConApp dc (map Type (tyConAppArgs scrut_ty)
2037 ++ varsToCoreExprs bndrs')
2039 LitAlt {} -> WARN( True, ptext (sLit "mkDupableAlt")
2040 <+> ppr case_bndr <+> ppr con )
2042 -- The case binder is alive but trivial, so why has
2043 -- it not been substituted away?
2045 used_bndrs' | isDeadBinder case_bndr = filter abstract_over bndrs'
2046 | otherwise = bndrs' ++ [case_bndr_w_unf]
2049 | isTyCoVar bndr = True -- Abstract over all type variables just in case
2050 | otherwise = not (isDeadBinder bndr)
2051 -- The deadness info on the new Ids is preserved by simplBinders
2053 ; (final_bndrs', final_args) -- Note [Join point abstraction]
2054 <- if (any isId used_bndrs')
2055 then return (used_bndrs', varsToCoreExprs used_bndrs')
2056 else do { rw_id <- newId (fsLit "w") realWorldStatePrimTy
2057 ; return ([rw_id], [Var realWorldPrimId]) }
2059 ; join_bndr <- newId (fsLit "$j") (mkPiTypes final_bndrs' rhs_ty')
2060 -- Note [Funky mkPiTypes]
2062 ; let -- We make the lambdas into one-shot-lambdas. The
2063 -- join point is sure to be applied at most once, and doing so
2064 -- prevents the body of the join point being floated out by
2065 -- the full laziness pass
2066 really_final_bndrs = map one_shot final_bndrs'
2067 one_shot v | isId v = setOneShotLambda v
2069 join_rhs = mkLams really_final_bndrs rhs'
2070 join_call = mkApps (Var join_bndr) final_args
2072 ; env' <- addPolyBind NotTopLevel env (NonRec join_bndr join_rhs)
2073 ; return (env', (con, bndrs', join_call)) }
2074 -- See Note [Duplicated env]
2077 Note [Case binders and join points]
2078 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2080 case (case .. ) of c {
2083 If we make a join point with c but not c# we get
2084 $j = \c -> ....c....
2086 But if later inlining scrutines the c, thus
2088 $j = \c -> ... case c of { I# y -> ... } ...
2090 we won't see that 'c' has already been scrutinised. This actually
2091 happens in the 'tabulate' function in wave4main, and makes a significant
2092 difference to allocation.
2094 An alternative plan is this:
2096 $j = \c# -> let c = I# c# in ...c....
2098 but that is bad if 'c' is *not* later scrutinised.
2100 So instead we do both: we pass 'c' and 'c#' , and record in c's inlining
2101 (an InlineRule) that it's really I# c#, thus
2103 $j = \c# -> \c[=I# c#] -> ...c....
2105 Absence analysis may later discard 'c'.
2107 NB: take great care when doing strictness analysis;
2108 see Note [Lamba-bound unfoldings] in DmdAnal.
2110 Also note that we can still end up passing stuff that isn't used. Before
2111 strictness analysis we have
2112 let $j x y c{=(x,y)} = (h c, ...)
2114 After strictness analysis we see that h is strict, we end up with
2115 let $j x y c{=(x,y)} = ($wh x y, ...)
2118 Note [Duplicated env]
2119 ~~~~~~~~~~~~~~~~~~~~~
2120 Some of the alternatives are simplified, but have not been turned into a join point
2121 So they *must* have an zapped subst-env. So we can't use completeNonRecX to
2122 bind the join point, because it might to do PostInlineUnconditionally, and
2123 we'd lose that when zapping the subst-env. We could have a per-alt subst-env,
2124 but zapping it (as we do in mkDupableCont, the Select case) is safe, and
2125 at worst delays the join-point inlining.
2127 Note [Small alternative rhs]
2128 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2129 It is worth checking for a small RHS because otherwise we
2130 get extra let bindings that may cause an extra iteration of the simplifier to
2131 inline back in place. Quite often the rhs is just a variable or constructor.
2132 The Ord instance of Maybe in PrelMaybe.lhs, for example, took several extra
2133 iterations because the version with the let bindings looked big, and so wasn't
2134 inlined, but after the join points had been inlined it looked smaller, and so
2137 NB: we have to check the size of rhs', not rhs.
2138 Duplicating a small InAlt might invalidate occurrence information
2139 However, if it *is* dupable, we return the *un* simplified alternative,
2140 because otherwise we'd need to pair it up with an empty subst-env....
2141 but we only have one env shared between all the alts.
2142 (Remember we must zap the subst-env before re-simplifying something).
2143 Rather than do this we simply agree to re-simplify the original (small) thing later.
2145 Note [Funky mkPiTypes]
2146 ~~~~~~~~~~~~~~~~~~~~~~
2147 Notice the funky mkPiTypes. If the contructor has existentials
2148 it's possible that the join point will be abstracted over
2149 type varaibles as well as term variables.
2150 Example: Suppose we have
2151 data T = forall t. C [t]
2153 case (case e of ...) of
2155 We get the join point
2156 let j :: forall t. [t] -> ...
2157 j = /\t \xs::[t] -> rhs
2159 case (case e of ...) of
2160 C t xs::[t] -> j t xs
2162 Note [Join point abstaction]
2163 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2164 If we try to lift a primitive-typed something out
2165 for let-binding-purposes, we will *caseify* it (!),
2166 with potentially-disastrous strictness results. So
2167 instead we turn it into a function: \v -> e
2168 where v::State# RealWorld#. The value passed to this function
2169 is realworld#, which generates (almost) no code.
2171 There's a slight infelicity here: we pass the overall
2172 case_bndr to all the join points if it's used in *any* RHS,
2173 because we don't know its usage in each RHS separately
2175 We used to say "&& isUnLiftedType rhs_ty'" here, but now
2176 we make the join point into a function whenever used_bndrs'
2177 is empty. This makes the join-point more CPR friendly.
2178 Consider: let j = if .. then I# 3 else I# 4
2179 in case .. of { A -> j; B -> j; C -> ... }
2181 Now CPR doesn't w/w j because it's a thunk, so
2182 that means that the enclosing function can't w/w either,
2183 which is a lose. Here's the example that happened in practice:
2184 kgmod :: Int -> Int -> Int
2185 kgmod x y = if x > 0 && y < 0 || x < 0 && y > 0
2189 I have seen a case alternative like this:
2191 It's a bit silly to add the realWorld dummy arg in this case, making
2194 (the \v alone is enough to make CPR happy) but I think it's rare
2196 Note [Duplicating StrictArg]
2197 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2198 The original plan had (where E is a big argument)
2200 ==> let $j = \a -> f E a
2203 But this is terrible! Here's an example:
2204 && E (case x of { T -> F; F -> T })
2205 Now, && is strict so we end up simplifying the case with
2206 an ArgOf continuation. If we let-bind it, we get
2207 let $j = \v -> && E v
2208 in simplExpr (case x of { T -> F; F -> T })
2210 And after simplifying more we get
2211 let $j = \v -> && E v
2212 in case x of { T -> $j F; F -> $j T }
2213 Which is a Very Bad Thing
2215 What we do now is this
2219 Now if the thing in the hole is a case expression (which is when
2220 we'll call mkDupableCont), we'll push the function call into the
2221 branches, which is what we want. Now RULES for f may fire, and
2222 call-pattern specialisation. Here's an example from Trac #3116
2225 _ -> Chunk p fpc (o+1) (l-1) bs')
2226 If we can push the call for 'go' inside the case, we get
2227 call-pattern specialisation for 'go', which is *crucial* for
2230 Here is the (&&) example:
2231 && E (case x of { T -> F; F -> T })
2233 case x of { T -> && a F; F -> && a T }
2237 * Arguments to f *after* the strict one are handled by
2238 the ApplyTo case of mkDupableCont. Eg
2241 * We can only do the let-binding of E because the function
2242 part of a StrictArg continuation is an explicit syntax
2243 tree. In earlier versions we represented it as a function
2244 (CoreExpr -> CoreEpxr) which we couldn't take apart.
2246 Do *not* duplicate StrictBind and StritArg continuations. We gain
2247 nothing by propagating them into the expressions, and we do lose a
2250 The desire not to duplicate is the entire reason that
2251 mkDupableCont returns a pair of continuations.
2253 Note [Duplicating StrictBind]
2254 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2255 Unlike StrictArg, there doesn't seem anything to gain from
2256 duplicating a StrictBind continuation, so we don't.
2258 The desire not to duplicate is the entire reason that
2259 mkDupableCont returns a pair of continuations.
2262 Note [Single-alternative cases]
2263 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2264 This case is just like the ArgOf case. Here's an example:
2268 case (case x of I# x' ->
2270 True -> I# (negate# x')
2271 False -> I# x') of y {
2273 Because the (case x) has only one alternative, we'll transform to
2275 case (case x' <# 0# of
2276 True -> I# (negate# x')
2277 False -> I# x') of y {
2279 But now we do *NOT* want to make a join point etc, giving
2281 let $j = \y -> MkT y
2283 True -> $j (I# (negate# x'))
2285 In this case the $j will inline again, but suppose there was a big
2286 strict computation enclosing the orginal call to MkT. Then, it won't
2287 "see" the MkT any more, because it's big and won't get duplicated.
2288 And, what is worse, nothing was gained by the case-of-case transform.
2290 So, in circumstances like these, we don't want to build join points
2291 and push the outer case into the branches of the inner one. Instead,
2292 don't duplicate the continuation.
2294 When should we use this strategy? We should not use it on *every*
2295 single-alternative case:
2296 e.g. case (case ....) of (a,b) -> (# a,b #)
2297 Here we must push the outer case into the inner one!
2300 * Match [(DEFAULT,_,_)], but in the common case of Int,
2301 the alternative-filling-in code turned the outer case into
2302 case (...) of y { I# _ -> MkT y }
2304 * Match on single alternative plus (not (isDeadBinder case_bndr))
2305 Rationale: pushing the case inwards won't eliminate the construction.
2306 But there's a risk of
2307 case (...) of y { (a,b) -> let z=(a,b) in ... }
2308 Now y looks dead, but it'll come alive again. Still, this
2309 seems like the best option at the moment.
2311 * Match on single alternative plus (all (isDeadBinder bndrs))
2312 Rationale: this is essentially seq.
2314 * Match when the rhs is *not* duplicable, and hence would lead to a
2315 join point. This catches the disaster-case above. We can test
2316 the *un-simplified* rhs, which is fine. It might get bigger or
2317 smaller after simplification; if it gets smaller, this case might
2318 fire next time round. NB also that we must test contIsDupable
2319 case_cont *too, because case_cont might be big!
2321 HOWEVER: I found that this version doesn't work well, because
2322 we can get let x = case (...) of { small } in ...case x...
2323 When x is inlined into its full context, we find that it was a bad
2324 idea to have pushed the outer case inside the (...) case.
2326 Note [Single-alternative-unlifted]
2327 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2328 Here's another single-alternative where we really want to do case-of-case:
2336 case y_s6X of tpl_s7m {
2337 M1.Mk1 ipv_s70 -> ipv_s70;
2338 M1.Mk2 ipv_s72 -> ipv_s72;
2344 case x_s74 of tpl_s7n {
2345 M1.Mk1 ipv_s77 -> ipv_s77;
2346 M1.Mk2 ipv_s79 -> ipv_s79;
2350 { __DEFAULT -> ==# [wild1_s7b wild_s7c];
2354 So the outer case is doing *nothing at all*, other than serving as a
2355 join-point. In this case we really want to do case-of-case and decide
2356 whether to use a real join point or just duplicate the continuation.
2358 Hence: check whether the case binder's type is unlifted, because then
2359 the outer case is *not* a seq.