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 = do { let rhs_env = rhs_se `setInScope` env
326 (tvs, body) = case collectTyBinders rhs of
327 (tvs, body) | not_lam body -> (tvs,body)
328 | otherwise -> ([], rhs)
329 not_lam (Lam _ _) = False
331 -- Do not do the "abstract tyyvar" thing if there's
332 -- a lambda inside, becuase it defeats eta-reduction
333 -- f = /\a. \x. g a x
336 ; (body_env, tvs') <- simplBinders rhs_env tvs
337 -- See Note [Floating and type abstraction] in SimplUtils
340 ; (body_env1, body1) <- simplExprF body_env body mkRhsStop
341 -- ANF-ise a constructor or PAP rhs
342 ; (body_env2, body2) <- prepareRhs top_lvl body_env1 bndr1 body1
345 <- if not (doFloatFromRhs top_lvl is_rec False body2 body_env2)
346 then -- No floating, revert to body1
347 do { rhs' <- mkLam env tvs' (wrapFloats body_env1 body1)
348 ; return (env, rhs') }
350 else if null tvs then -- Simple floating
351 do { tick LetFloatFromLet
352 ; return (addFloats env body_env2, body2) }
354 else -- Do type-abstraction first
355 do { tick LetFloatFromLet
356 ; (poly_binds, body3) <- abstractFloats tvs' body_env2 body2
357 ; rhs' <- mkLam env tvs' body3
358 ; env' <- foldlM (addPolyBind top_lvl) env poly_binds
359 ; return (env', rhs') }
361 ; completeBind env' top_lvl bndr bndr1 rhs' }
364 A specialised variant of simplNonRec used when the RHS is already simplified,
365 notably in knownCon. It uses case-binding where necessary.
368 simplNonRecX :: SimplEnv
369 -> InId -- Old binder
370 -> OutExpr -- Simplified RHS
373 simplNonRecX env bndr new_rhs
374 | isDeadBinder bndr -- Not uncommon; e.g. case (a,b) of b { (p,q) -> p }
375 = return env -- Here b is dead, and we avoid creating
376 | otherwise -- the binding b = (a,b)
377 = do { (env', bndr') <- simplBinder env bndr
378 ; completeNonRecX NotTopLevel env' (isStrictId bndr) bndr bndr' new_rhs }
379 -- simplNonRecX is only used for NotTopLevel things
381 completeNonRecX :: TopLevelFlag -> SimplEnv
383 -> InId -- Old binder
384 -> OutId -- New binder
385 -> OutExpr -- Simplified RHS
388 completeNonRecX top_lvl env is_strict old_bndr new_bndr new_rhs
389 = do { (env1, rhs1) <- prepareRhs top_lvl (zapFloats env) new_bndr new_rhs
391 if doFloatFromRhs NotTopLevel NonRecursive is_strict rhs1 env1
392 then do { tick LetFloatFromLet
393 ; return (addFloats env env1, rhs1) } -- Add the floats to the main env
394 else return (env, wrapFloats env1 rhs1) -- Wrap the floats around the RHS
395 ; completeBind env2 NotTopLevel old_bndr new_bndr rhs2 }
398 {- No, no, no! Do not try preInlineUnconditionally in completeNonRecX
399 Doing so risks exponential behaviour, because new_rhs has been simplified once already
400 In the cases described by the folowing commment, postInlineUnconditionally will
401 catch many of the relevant cases.
402 -- This happens; for example, the case_bndr during case of
403 -- known constructor: case (a,b) of x { (p,q) -> ... }
404 -- Here x isn't mentioned in the RHS, so we don't want to
405 -- create the (dead) let-binding let x = (a,b) in ...
407 -- Similarly, single occurrences can be inlined vigourously
408 -- e.g. case (f x, g y) of (a,b) -> ....
409 -- If a,b occur once we can avoid constructing the let binding for them.
411 Furthermore in the case-binding case preInlineUnconditionally risks extra thunks
412 -- Consider case I# (quotInt# x y) of
413 -- I# v -> let w = J# v in ...
414 -- If we gaily inline (quotInt# x y) for v, we end up building an
416 -- let w = J# (quotInt# x y) in ...
417 -- because quotInt# can fail.
419 | preInlineUnconditionally env NotTopLevel bndr new_rhs
420 = thing_inside (extendIdSubst env bndr (DoneEx new_rhs))
423 ----------------------------------
424 prepareRhs takes a putative RHS, checks whether it's a PAP or
425 constructor application and, if so, converts it to ANF, so that the
426 resulting thing can be inlined more easily. Thus
433 We also want to deal well cases like this
434 v = (f e1 `cast` co) e2
435 Here we want to make e1,e2 trivial and get
436 x1 = e1; x2 = e2; v = (f x1 `cast` co) v2
437 That's what the 'go' loop in prepareRhs does
440 prepareRhs :: TopLevelFlag -> SimplEnv -> OutId -> OutExpr -> SimplM (SimplEnv, OutExpr)
441 -- Adds new floats to the env iff that allows us to return a good RHS
442 prepareRhs top_lvl env id (Cast rhs co) -- Note [Float coercions]
443 | (ty1, _ty2) <- coercionKind co -- Do *not* do this if rhs has an unlifted type
444 , not (isUnLiftedType ty1) -- see Note [Float coercions (unlifted)]
445 = do { (env', rhs') <- makeTrivialWithInfo top_lvl env sanitised_info rhs
446 ; return (env', Cast rhs' co) }
448 sanitised_info = vanillaIdInfo `setStrictnessInfo` strictnessInfo info
449 `setDemandInfo` demandInfo info
452 prepareRhs top_lvl env0 _ rhs0
453 = do { (_is_exp, env1, rhs1) <- go 0 env0 rhs0
454 ; return (env1, rhs1) }
456 go n_val_args env (Cast rhs co)
457 = do { (is_exp, env', rhs') <- go n_val_args env rhs
458 ; return (is_exp, env', Cast rhs' co) }
459 go n_val_args env (App fun (Type ty))
460 = do { (is_exp, env', rhs') <- go n_val_args env fun
461 ; return (is_exp, env', App rhs' (Type ty)) }
462 go n_val_args env (App fun arg)
463 = do { (is_exp, env', fun') <- go (n_val_args+1) env fun
465 True -> do { (env'', arg') <- makeTrivial top_lvl env' arg
466 ; return (True, env'', App fun' arg') }
467 False -> return (False, env, App fun arg) }
468 go n_val_args env (Var fun)
469 = return (is_exp, env, Var fun)
471 is_exp = isExpandableApp fun n_val_args -- The fun a constructor or PAP
472 -- See Note [CONLIKE pragma] in BasicTypes
473 -- The definition of is_exp should match that in
474 -- OccurAnal.occAnalApp
477 = return (False, env, other)
481 Note [Float coercions]
482 ~~~~~~~~~~~~~~~~~~~~~~
483 When we find the binding
485 we'd like to transform it to
487 x = x `cast` co -- A trivial binding
488 There's a chance that e will be a constructor application or function, or something
489 like that, so moving the coerion to the usage site may well cancel the coersions
490 and lead to further optimisation. Example:
493 data instance T Int = T Int
495 foo :: Int -> Int -> Int
500 go n = case x of { T m -> go (n-m) }
501 -- This case should optimise
503 Note [Preserve strictness when floating coercions]
504 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
505 In the Note [Float coercions] transformation, keep the strictness info.
507 f = e `cast` co -- f has strictness SSL
509 f' = e -- f' also has strictness SSL
510 f = f' `cast` co -- f still has strictness SSL
512 Its not wrong to drop it on the floor, but better to keep it.
514 Note [Float coercions (unlifted)]
515 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
516 BUT don't do [Float coercions] if 'e' has an unlifted type.
519 foo :: Int = (error (# Int,Int #) "urk")
520 `cast` CoUnsafe (# Int,Int #) Int
522 If do the makeTrivial thing to the error call, we'll get
523 foo = case error (# Int,Int #) "urk" of v -> v `cast` ...
524 But 'v' isn't in scope!
526 These strange casts can happen as a result of case-of-case
527 bar = case (case x of { T -> (# 2,3 #); F -> error "urk" }) of
532 makeTrivial :: TopLevelFlag -> SimplEnv -> OutExpr -> SimplM (SimplEnv, OutExpr)
533 -- Binds the expression to a variable, if it's not trivial, returning the variable
534 makeTrivial top_lvl env expr = makeTrivialWithInfo top_lvl env vanillaIdInfo expr
536 makeTrivialWithInfo :: TopLevelFlag -> SimplEnv -> IdInfo
537 -> OutExpr -> SimplM (SimplEnv, OutExpr)
538 -- Propagate strictness and demand info to the new binder
539 -- Note [Preserve strictness when floating coercions]
540 -- Returned SimplEnv has same substitution as incoming one
541 makeTrivialWithInfo top_lvl env info expr
542 | exprIsTrivial expr -- Already trivial
543 || not (bindingOk top_lvl expr expr_ty) -- Cannot trivialise
544 -- See Note [Cannot trivialise]
546 | otherwise -- See Note [Take care] below
547 = do { uniq <- getUniqueM
548 ; let name = mkSystemVarName uniq (fsLit "a")
549 var = mkLocalIdWithInfo name expr_ty info
550 ; env' <- completeNonRecX top_lvl env False var var expr
551 ; expr' <- simplVar env' var
552 ; return (env', expr') }
553 -- The simplVar is needed becase we're constructing a new binding
555 -- And if rhs is of form (rhs1 |> co), then we might get
558 -- and now a's RHS is trivial and can be substituted out, and that
559 -- is what completeNonRecX will do
560 -- To put it another way, it's as if we'd simplified
561 -- let var = e in var
563 expr_ty = exprType expr
565 bindingOk :: TopLevelFlag -> CoreExpr -> Type -> Bool
566 -- True iff we can have a binding of this expression at this level
567 -- Precondition: the type is the type of the expression
568 bindingOk top_lvl _ expr_ty
569 | isTopLevel top_lvl = not (isUnLiftedType expr_ty)
573 Note [Cannot trivialise]
574 ~~~~~~~~~~~~~~~~~~~~~~~~
581 Then we can't ANF-ise foo, even though we'd like to, because
582 we can't make a top-level binding for the Addr# (f 3). And if
583 so we don't want to turn it into
584 foo = let x = f 3 in Bar x
585 because we'll just end up inlining x back, and that makes the
586 simplifier loop. Better not to ANF-ise it at all.
588 A case in point is literal strings (a MachStr is not regarded as
593 We don't want to ANF-ise this.
595 %************************************************************************
597 \subsection{Completing a lazy binding}
599 %************************************************************************
602 * deals only with Ids, not TyVars
603 * takes an already-simplified binder and RHS
604 * is used for both recursive and non-recursive bindings
605 * is used for both top-level and non-top-level bindings
607 It does the following:
608 - tries discarding a dead binding
609 - tries PostInlineUnconditionally
610 - add unfolding [this is the only place we add an unfolding]
613 It does *not* attempt to do let-to-case. Why? Because it is used for
614 - top-level bindings (when let-to-case is impossible)
615 - many situations where the "rhs" is known to be a WHNF
616 (so let-to-case is inappropriate).
618 Nor does it do the atomic-argument thing
621 completeBind :: SimplEnv
622 -> TopLevelFlag -- Flag stuck into unfolding
623 -> InId -- Old binder
624 -> OutId -> OutExpr -- New binder and RHS
626 -- completeBind may choose to do its work
627 -- * by extending the substitution (e.g. let x = y in ...)
628 -- * or by adding to the floats in the envt
630 completeBind env top_lvl old_bndr new_bndr new_rhs
631 = do { let old_info = idInfo old_bndr
632 old_unf = unfoldingInfo old_info
633 occ_info = occInfo old_info
635 ; new_unfolding <- simplUnfolding env top_lvl old_bndr occ_info new_rhs old_unf
637 ; if postInlineUnconditionally env top_lvl new_bndr occ_info new_rhs new_unfolding
638 -- Inline and discard the binding
639 then do { tick (PostInlineUnconditionally old_bndr)
640 ; -- pprTrace "postInlineUnconditionally" (ppr old_bndr <+> equals <+> ppr new_rhs) $
641 return (extendIdSubst env old_bndr (DoneEx new_rhs)) }
642 -- Use the substitution to make quite, quite sure that the
643 -- substitution will happen, since we are going to discard the binding
645 else return (addNonRecWithUnf env new_bndr new_rhs new_unfolding) }
647 ------------------------------
648 addPolyBind :: TopLevelFlag -> SimplEnv -> OutBind -> SimplM SimplEnv
649 -- Add a new binding to the environment, complete with its unfolding
650 -- but *do not* do postInlineUnconditionally, because we have already
651 -- processed some of the scope of the binding
652 -- We still want the unfolding though. Consider
654 -- x = /\a. let y = ... in Just y
656 -- Then we float the y-binding out (via abstractFloats and addPolyBind)
657 -- but 'x' may well then be inlined in 'body' in which case we'd like the
658 -- opportunity to inline 'y' too.
660 addPolyBind top_lvl env (NonRec poly_id rhs)
661 = do { unfolding <- simplUnfolding env top_lvl poly_id NoOccInfo rhs noUnfolding
662 -- Assumes that poly_id did not have an INLINE prag
663 -- which is perhaps wrong. ToDo: think about this
664 ; return (addNonRecWithUnf env poly_id rhs unfolding) }
666 addPolyBind _ env bind@(Rec _) = return (extendFloats env bind)
667 -- Hack: letrecs are more awkward, so we extend "by steam"
668 -- without adding unfoldings etc. At worst this leads to
669 -- more simplifier iterations
671 ------------------------------
672 addNonRecWithUnf :: SimplEnv
673 -> OutId -> OutExpr -- New binder and RHS
674 -> Unfolding -- New unfolding
676 addNonRecWithUnf env new_bndr new_rhs new_unfolding
677 = let new_arity = exprArity new_rhs
678 old_arity = idArity new_bndr
679 info1 = idInfo new_bndr `setArityInfo` new_arity
681 -- Unfolding info: Note [Setting the new unfolding]
682 info2 = info1 `setUnfoldingInfo` new_unfolding
684 -- Demand info: Note [Setting the demand info]
685 info3 | isEvaldUnfolding new_unfolding = zapDemandInfo info2 `orElse` info2
688 final_id = new_bndr `setIdInfo` info3
689 dmd_arity = length $ fst $ splitStrictSig $ idStrictness new_bndr
691 ASSERT( isId new_bndr )
692 WARN( new_arity < old_arity || new_arity < dmd_arity,
693 (ptext (sLit "Arity decrease:") <+> (ppr final_id <+> ppr old_arity
694 <+> ppr new_arity <+> ppr dmd_arity) $$ ppr new_rhs) )
695 -- Note [Arity decrease]
697 final_id `seq` -- This seq forces the Id, and hence its IdInfo,
698 -- and hence any inner substitutions
699 -- pprTrace "Binding" (ppr final_id <+> ppr unfolding) $
700 addNonRec env final_id new_rhs
701 -- The addNonRec adds it to the in-scope set too
703 ------------------------------
704 simplUnfolding :: SimplEnv-> TopLevelFlag
706 -> OccInfo -> OutExpr
707 -> Unfolding -> SimplM Unfolding
708 -- Note [Setting the new unfolding]
709 simplUnfolding env _ _ _ _ (DFunUnfolding ar con ops)
710 = return (DFunUnfolding ar con ops')
712 ops' = map (substExpr (text "simplUnfolding") env) ops
714 simplUnfolding env top_lvl id _ _
715 (CoreUnfolding { uf_tmpl = expr, uf_arity = arity
716 , uf_src = src, uf_guidance = guide })
718 = do { expr' <- simplExpr rule_env expr
719 ; let src' = CoreSubst.substUnfoldingSource (mkCoreSubst (text "inline-unf") env) src
720 ; return (mkCoreUnfolding (isTopLevel top_lvl) src' expr' arity guide) }
721 -- See Note [Top-level flag on inline rules] in CoreUnfold
723 act = idInlineActivation id
724 rule_env = updMode (updModeForInlineRules act) env
725 -- See Note [Simplifying gently inside InlineRules] in SimplUtils
727 simplUnfolding _ top_lvl id _occ_info new_rhs _
728 = return (mkUnfolding InlineRhs (isTopLevel top_lvl) (isBottomingId id) new_rhs)
729 -- We make an unfolding *even for loop-breakers*.
730 -- Reason: (a) It might be useful to know that they are WHNF
731 -- (b) In TidyPgm we currently assume that, if we want to
732 -- expose the unfolding then indeed we *have* an unfolding
733 -- to expose. (We could instead use the RHS, but currently
734 -- we don't.) The simple thing is always to have one.
737 Note [Arity decrease]
738 ~~~~~~~~~~~~~~~~~~~~~
739 Generally speaking the arity of a binding should not decrease. But it *can*
740 legitimately happen becuase of RULES. Eg
742 where g has arity 2, will have arity 2. But if there's a rewrite rule
744 where h has arity 1, then f's arity will decrease. Here's a real-life example,
745 which is in the output of Specialise:
748 $dm {Arity 2} = \d.\x. op d
749 {-# RULES forall d. $dm Int d = $s$dm #-}
751 dInt = MkD .... opInt ...
752 opInt {Arity 1} = $dm dInt
754 $s$dm {Arity 0} = \x. op dInt }
756 Here opInt has arity 1; but when we apply the rule its arity drops to 0.
757 That's why Specialise goes to a little trouble to pin the right arity
758 on specialised functions too.
760 Note [Setting the new unfolding]
761 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
762 * If there's an INLINE pragma, we simplify the RHS gently. Maybe we
763 should do nothing at all, but simplifying gently might get rid of
766 * If not, we make an unfolding from the new RHS. But *only* for
767 non-loop-breakers. Making loop breakers not have an unfolding at all
768 means that we can avoid tests in exprIsConApp, for example. This is
769 important: if exprIsConApp says 'yes' for a recursive thing, then we
770 can get into an infinite loop
772 If there's an InlineRule on a loop breaker, we hang on to the inlining.
773 It's pretty dodgy, but the user did say 'INLINE'. May need to revisit
776 Note [Setting the demand info]
777 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
778 If the unfolding is a value, the demand info may
779 go pear-shaped, so we nuke it. Example:
781 case x of (p,q) -> h p q x
782 Here x is certainly demanded. But after we've nuked
783 the case, we'll get just
784 let x = (a,b) in h a b x
785 and now x is not demanded (I'm assuming h is lazy)
786 This really happens. Similarly
787 let f = \x -> e in ...f..f...
788 After inlining f at some of its call sites the original binding may
789 (for example) be no longer strictly demanded.
790 The solution here is a bit ad hoc...
793 %************************************************************************
795 \subsection[Simplify-simplExpr]{The main function: simplExpr}
797 %************************************************************************
799 The reason for this OutExprStuff stuff is that we want to float *after*
800 simplifying a RHS, not before. If we do so naively we get quadratic
801 behaviour as things float out.
803 To see why it's important to do it after, consider this (real) example:
817 a -- Can't inline a this round, cos it appears twice
821 Each of the ==> steps is a round of simplification. We'd save a
822 whole round if we float first. This can cascade. Consider
827 let f = let d1 = ..d.. in \y -> e
831 in \x -> ...(\y ->e)...
833 Only in this second round can the \y be applied, and it
834 might do the same again.
838 simplExpr :: SimplEnv -> CoreExpr -> SimplM CoreExpr
839 simplExpr env expr = simplExprC env expr mkBoringStop
841 simplExprC :: SimplEnv -> CoreExpr -> SimplCont -> SimplM CoreExpr
842 -- Simplify an expression, given a continuation
843 simplExprC env expr cont
844 = -- pprTrace "simplExprC" (ppr expr $$ ppr cont {- $$ ppr (seIdSubst env) -} $$ ppr (seFloats env) ) $
845 do { (env', expr') <- simplExprF (zapFloats env) expr cont
846 ; -- pprTrace "simplExprC ret" (ppr expr $$ ppr expr') $
847 -- pprTrace "simplExprC ret3" (ppr (seInScope env')) $
848 -- pprTrace "simplExprC ret4" (ppr (seFloats env')) $
849 return (wrapFloats env' expr') }
851 --------------------------------------------------
852 simplExprF :: SimplEnv -> InExpr -> SimplCont
853 -> SimplM (SimplEnv, OutExpr)
855 simplExprF env e cont
856 = -- pprTrace "simplExprF" (ppr e $$ ppr cont $$ ppr (seTvSubst env) $$ ppr (seIdSubst env) {- $$ ppr (seFloats env) -} ) $
857 simplExprF' env e cont
859 simplExprF' :: SimplEnv -> InExpr -> SimplCont
860 -> SimplM (SimplEnv, OutExpr)
861 simplExprF' env (Var v) cont = simplVarF env v cont
862 simplExprF' env (Lit lit) cont = rebuild env (Lit lit) cont
863 simplExprF' env (Note n expr) cont = simplNote env n expr cont
864 simplExprF' env (Cast body co) cont = simplCast env body co cont
865 simplExprF' env (App fun arg) cont = simplExprF env fun $
866 ApplyTo NoDup arg env cont
868 simplExprF' env expr@(Lam _ _) cont
869 = simplLam env (map zap bndrs) body cont
870 -- The main issue here is under-saturated lambdas
871 -- (\x1. \x2. e) arg1
872 -- Here x1 might have "occurs-once" occ-info, because occ-info
873 -- is computed assuming that a group of lambdas is applied
874 -- all at once. If there are too few args, we must zap the
877 n_args = countArgs cont
878 n_params = length bndrs
879 (bndrs, body) = collectBinders expr
880 zap | n_args >= n_params = \b -> b
881 | otherwise = \b -> if isTyCoVar b then b
883 -- NB: we count all the args incl type args
884 -- so we must count all the binders (incl type lambdas)
886 simplExprF' env (Type ty) cont
887 = ASSERT( contIsRhsOrArg cont )
888 do { ty' <- simplCoercion env ty
889 ; rebuild env (Type ty') cont }
891 simplExprF' env (Case scrut bndr _ alts) cont
892 | not (switchIsOn (getSwitchChecker env) NoCaseOfCase)
893 = -- Simplify the scrutinee with a Select continuation
894 simplExprF env scrut (Select NoDup bndr alts env cont)
897 = -- If case-of-case is off, simply simplify the case expression
898 -- in a vanilla Stop context, and rebuild the result around it
899 do { case_expr' <- simplExprC env scrut case_cont
900 ; rebuild env case_expr' cont }
902 case_cont = Select NoDup bndr alts env mkBoringStop
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 _ arg se cont -> do { arg' <- simplExpr (se `setInScope` env) arg
955 ; rebuild env (App expr arg') cont }
959 %************************************************************************
963 %************************************************************************
966 simplCast :: SimplEnv -> InExpr -> Coercion -> SimplCont
967 -> SimplM (SimplEnv, OutExpr)
968 simplCast env body co0 cont0
969 = do { co1 <- simplCoercion env co0
970 ; simplExprF env body (addCoerce co1 cont0) }
972 addCoerce co cont = add_coerce co (coercionKind co) cont
974 add_coerce _co (s1, k1) cont -- co :: ty~ty
975 | s1 `coreEqType` k1 = cont -- is a no-op
977 add_coerce co1 (s1, _k2) (CoerceIt co2 cont)
978 | (_l1, t1) <- coercionKind co2
979 -- e |> (g1 :: S1~L) |> (g2 :: L~T1)
982 -- e |> (g1 . g2 :: S1~T1) otherwise
984 -- For example, in the initial form of a worker
985 -- we may find (coerce T (coerce S (\x.e))) y
986 -- and we'd like it to simplify to e[y/x] in one round
988 , s1 `coreEqType` t1 = cont -- The coerces cancel out
989 | otherwise = CoerceIt (mkTransCoercion co1 co2) cont
991 add_coerce co (s1s2, _t1t2) (ApplyTo dup (Type arg_ty) arg_se cont)
992 -- (f |> g) ty ---> (f ty) |> (g @ ty)
993 -- This implements the PushT and PushC rules from the paper
994 | Just (tyvar,_) <- splitForAllTy_maybe s1s2
996 (new_arg_ty, new_cast)
997 | isCoVar tyvar = (new_arg_co, mkCselRCoercion co) -- PushC rule
998 | otherwise = (ty', mkInstCoercion co ty') -- PushT rule
1000 ApplyTo dup (Type new_arg_ty) (zapSubstEnv arg_se) (addCoerce new_cast cont)
1002 ty' = substTy (arg_se `setInScope` env) arg_ty
1003 new_arg_co = mkCsel1Coercion co `mkTransCoercion`
1004 ty' `mkTransCoercion`
1005 mkSymCoercion (mkCsel2Coercion co)
1007 add_coerce co (s1s2, _t1t2) (ApplyTo dup arg arg_se cont)
1008 | not (isTypeArg arg) -- This implements the Push rule from the paper
1009 , isFunTy s1s2 -- t1t2 must be a function type, becuase it's applied
1010 -- (e |> (g :: s1s2 ~ t1->t2)) f
1012 -- (e (f |> (arg g :: t1~s1))
1013 -- |> (res g :: s2->t2)
1015 -- t1t2 must be a function type, t1->t2, because it's applied
1016 -- to something but s1s2 might conceivably not be
1018 -- When we build the ApplyTo we can't mix the out-types
1019 -- with the InExpr in the argument, so we simply substitute
1020 -- to make it all consistent. It's a bit messy.
1021 -- But it isn't a common case.
1023 -- Example of use: Trac #995
1024 = ApplyTo dup new_arg (zapSubstEnv arg_se) (addCoerce co2 cont)
1026 -- we split coercion t1->t2 ~ s1->s2 into t1 ~ s1 and
1027 -- t2 ~ s2 with left and right on the curried form:
1028 -- (->) t1 t2 ~ (->) s1 s2
1029 [co1, co2] = decomposeCo 2 co
1030 new_arg = mkCoerce (mkSymCoercion co1) arg'
1031 arg' = substExpr (text "move-cast") (arg_se `setInScope` env) arg
1033 add_coerce co _ cont = CoerceIt co cont
1037 %************************************************************************
1039 \subsection{Lambdas}
1041 %************************************************************************
1044 simplLam :: SimplEnv -> [InId] -> InExpr -> SimplCont
1045 -> SimplM (SimplEnv, OutExpr)
1047 simplLam env [] body cont = simplExprF env body cont
1050 simplLam env (bndr:bndrs) body (ApplyTo _ arg arg_se cont)
1051 = do { tick (BetaReduction bndr)
1052 ; simplNonRecE env bndr (arg, arg_se) (bndrs, body) cont }
1054 -- Not enough args, so there are real lambdas left to put in the result
1055 simplLam env bndrs body cont
1056 = do { (env', bndrs') <- simplLamBndrs env bndrs
1057 ; body' <- simplExpr env' body
1058 ; new_lam <- mkLam env' bndrs' body'
1059 ; rebuild env' new_lam cont }
1062 simplNonRecE :: SimplEnv
1063 -> InBndr -- The binder
1064 -> (InExpr, SimplEnv) -- Rhs of binding (or arg of lambda)
1065 -> ([InBndr], InExpr) -- Body of the let/lambda
1068 -> SimplM (SimplEnv, OutExpr)
1070 -- simplNonRecE is used for
1071 -- * non-top-level non-recursive lets in expressions
1074 -- It deals with strict bindings, via the StrictBind continuation,
1075 -- which may abort the whole process
1077 -- The "body" of the binding comes as a pair of ([InId],InExpr)
1078 -- representing a lambda; so we recurse back to simplLam
1079 -- Why? Because of the binder-occ-info-zapping done before
1080 -- the call to simplLam in simplExprF (Lam ...)
1082 -- First deal with type applications and type lets
1083 -- (/\a. e) (Type ty) and (let a = Type ty in e)
1084 simplNonRecE env bndr (Type ty_arg, rhs_se) (bndrs, body) cont
1085 = ASSERT( isTyCoVar bndr )
1086 do { ty_arg' <- simplType (rhs_se `setInScope` env) ty_arg
1087 ; simplLam (extendTvSubst env bndr ty_arg') bndrs body cont }
1089 simplNonRecE env bndr (rhs, rhs_se) (bndrs, body) cont
1090 | preInlineUnconditionally env NotTopLevel bndr rhs
1091 = do { tick (PreInlineUnconditionally bndr)
1092 ; simplLam (extendIdSubst env bndr (mkContEx rhs_se rhs)) bndrs body cont }
1095 = do { simplExprF (rhs_se `setFloats` env) rhs
1096 (StrictBind bndr bndrs body env cont) }
1099 = ASSERT( not (isTyCoVar bndr) )
1100 do { (env1, bndr1) <- simplNonRecBndr env bndr
1101 ; let (env2, bndr2) = addBndrRules env1 bndr bndr1
1102 ; env3 <- simplLazyBind env2 NotTopLevel NonRecursive bndr bndr2 rhs rhs_se
1103 ; simplLam env3 bndrs body cont }
1107 %************************************************************************
1111 %************************************************************************
1114 -- Hack alert: we only distinguish subsumed cost centre stacks for the
1115 -- purposes of inlining. All other CCCSs are mapped to currentCCS.
1116 simplNote :: SimplEnv -> Note -> CoreExpr -> SimplCont
1117 -> SimplM (SimplEnv, OutExpr)
1118 simplNote env (SCC cc) e cont
1119 | pushCCisNop cc (getEnclosingCC env) -- scc "f" (...(scc "f" e)...)
1120 = simplExprF env e cont -- ==> scc "f" (...e...)
1122 = do { e' <- simplExpr (setEnclosingCC env currentCCS) e
1123 ; rebuild env (mkSCC cc e') cont }
1125 simplNote env (CoreNote s) e cont
1126 = do { e' <- simplExpr env e
1127 ; rebuild env (Note (CoreNote s) e') cont }
1131 %************************************************************************
1135 %************************************************************************
1138 simplVar :: SimplEnv -> InVar -> SimplM OutExpr
1139 -- Look up an InVar in the environment
1142 = return (Type (substTyVar env var))
1144 = case substId env var of
1145 DoneId var1 -> return (Var var1)
1146 DoneEx e -> return e
1147 ContEx tvs ids e -> simplExpr (setSubstEnv env tvs ids) e
1149 simplVarF :: SimplEnv -> InId -> SimplCont -> SimplM (SimplEnv, OutExpr)
1150 simplVarF env var cont
1151 = case substId env var of
1152 DoneEx e -> simplExprF (zapSubstEnv env) e cont
1153 ContEx tvs ids e -> simplExprF (setSubstEnv env tvs ids) e cont
1154 DoneId var1 -> completeCall env var1 cont
1155 -- Note [zapSubstEnv]
1156 -- The template is already simplified, so don't re-substitute.
1157 -- This is VITAL. Consider
1159 -- let y = \z -> ...x... in
1161 -- We'll clone the inner \x, adding x->x' in the id_subst
1162 -- Then when we inline y, we must *not* replace x by x' in
1163 -- the inlined copy!!
1165 ---------------------------------------------------------
1166 -- Dealing with a call site
1168 completeCall :: SimplEnv -> Id -> SimplCont -> SimplM (SimplEnv, OutExpr)
1169 completeCall env var cont
1170 = do { ------------- Try inlining ----------------
1171 dflags <- getDOptsSmpl
1172 ; let (lone_variable, arg_infos, call_cont) = contArgs cont
1173 -- The args are OutExprs, obtained by *lazily* substituting
1174 -- in the args found in cont. These args are only examined
1175 -- to limited depth (unless a rule fires). But we must do
1176 -- the substitution; rule matching on un-simplified args would
1179 n_val_args = length arg_infos
1180 interesting_cont = interestingCallContext call_cont
1181 unfolding = activeUnfolding env var
1182 maybe_inline = callSiteInline dflags var unfolding
1183 lone_variable arg_infos interesting_cont
1184 ; case maybe_inline of {
1185 Just expr -- There is an inlining!
1186 -> do { tick (UnfoldingDone var)
1187 ; trace_inline dflags expr cont $
1188 simplExprF (zapSubstEnv env) expr cont }
1190 ; Nothing -> do -- No inlining!
1192 { rule_base <- getSimplRules
1193 ; let info = mkArgInfo var (getRules rule_base var) n_val_args call_cont
1194 ; rebuildCall env info cont
1197 trace_inline dflags unfolding cont stuff
1198 | not (dopt Opt_D_dump_inlinings dflags) = stuff
1199 | not (dopt Opt_D_verbose_core2core dflags)
1200 = if isExternalName (idName var) then
1201 pprTrace "Inlining done:" (ppr var) stuff
1204 = pprTrace ("Inlining done: " ++ showSDoc (ppr var))
1205 (vcat [text "Inlined fn: " <+> nest 2 (ppr unfolding),
1206 text "Cont: " <+> ppr cont])
1209 rebuildCall :: SimplEnv
1212 -> SimplM (SimplEnv, OutExpr)
1213 rebuildCall env (ArgInfo { ai_fun = fun, ai_args = rev_args, ai_strs = [] }) cont
1214 -- When we run out of strictness args, it means
1215 -- that the call is definitely bottom; see SimplUtils.mkArgInfo
1216 -- Then we want to discard the entire strict continuation. E.g.
1217 -- * case (error "hello") of { ... }
1218 -- * (error "Hello") arg
1219 -- * f (error "Hello") where f is strict
1221 -- Then, especially in the first of these cases, we'd like to discard
1222 -- the continuation, leaving just the bottoming expression. But the
1223 -- type might not be right, so we may have to add a coerce.
1224 | not (contIsTrivial cont) -- Only do this if there is a non-trivial
1225 = return (env, mk_coerce res) -- contination to discard, else we do it
1226 where -- again and again!
1227 res = mkApps (Var fun) (reverse rev_args)
1228 res_ty = exprType res
1229 cont_ty = contResultType env res_ty cont
1230 co = mkUnsafeCoercion res_ty cont_ty
1231 mk_coerce expr | cont_ty `coreEqType` res_ty = expr
1232 | otherwise = mkCoerce co expr
1234 rebuildCall env info (ApplyTo _ (Type arg_ty) se cont)
1235 = do { ty' <- simplCoercion (se `setInScope` env) arg_ty
1236 ; rebuildCall env (info `addArgTo` Type ty') cont }
1238 rebuildCall env info@(ArgInfo { ai_encl = encl_rules
1239 , ai_strs = str:strs, ai_discs = disc:discs })
1240 (ApplyTo _ arg arg_se cont)
1241 | str -- Strict argument
1242 = -- pprTrace "Strict Arg" (ppr arg $$ ppr (seIdSubst env) $$ ppr (seInScope env)) $
1243 simplExprF (arg_se `setFloats` env) arg
1244 (StrictArg info' cci cont)
1247 | otherwise -- Lazy argument
1248 -- DO NOT float anything outside, hence simplExprC
1249 -- There is no benefit (unlike in a let-binding), and we'd
1250 -- have to be very careful about bogus strictness through
1251 -- floating a demanded let.
1252 = do { arg' <- simplExprC (arg_se `setInScope` env) arg
1254 ; rebuildCall env (addArgTo info' arg') cont }
1256 info' = info { ai_strs = strs, ai_discs = discs }
1257 cci | encl_rules || disc > 0 = ArgCtxt encl_rules -- Be keener here
1258 | otherwise = BoringCtxt -- Nothing interesting
1260 rebuildCall env (ArgInfo { ai_fun = fun, ai_args = rev_args, ai_rules = rules }) cont
1261 = do { -- We've accumulated a simplified call in <fun,rev_args>
1262 -- so try rewrite rules; see Note [RULEs apply to simplified arguments]
1263 -- See also Note [Rules for recursive functions]
1264 ; let args = reverse rev_args
1265 env' = zapSubstEnv env
1266 ; mb_rule <- tryRules env rules fun args cont
1268 Just (n_args, rule_rhs) -> simplExprF env' rule_rhs $
1269 pushArgs env' (drop n_args args) cont ;
1270 -- n_args says how many args the rule consumed
1271 ; Nothing -> rebuild env (mkApps (Var fun) args) cont -- No rules
1275 Note [RULES apply to simplified arguments]
1276 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1277 It's very desirable to try RULES once the arguments have been simplified, because
1278 doing so ensures that rule cascades work in one pass. Consider
1279 {-# RULES g (h x) = k x
1282 Then we want to rewrite (g (h x)) to (k x) and only then try f's rules. If
1283 we match f's rules against the un-simplified RHS, it won't match. This
1284 makes a particularly big difference when superclass selectors are involved:
1285 op ($p1 ($p2 (df d)))
1286 We want all this to unravel in one sweeep.
1290 This part of the simplifier may break the no-shadowing invariant
1292 f (...(\a -> e)...) (case y of (a,b) -> e')
1293 where f is strict in its second arg
1294 If we simplify the innermost one first we get (...(\a -> e)...)
1295 Simplifying the second arg makes us float the case out, so we end up with
1296 case y of (a,b) -> f (...(\a -> e)...) e'
1297 So the output does not have the no-shadowing invariant. However, there is
1298 no danger of getting name-capture, because when the first arg was simplified
1299 we used an in-scope set that at least mentioned all the variables free in its
1300 static environment, and that is enough.
1302 We can't just do innermost first, or we'd end up with a dual problem:
1303 case x of (a,b) -> f e (...(\a -> e')...)
1305 I spent hours trying to recover the no-shadowing invariant, but I just could
1306 not think of an elegant way to do it. The simplifier is already knee-deep in
1307 continuations. We have to keep the right in-scope set around; AND we have
1308 to get the effect that finding (error "foo") in a strict arg position will
1309 discard the entire application and replace it with (error "foo"). Getting
1310 all this at once is TOO HARD!
1313 %************************************************************************
1317 %************************************************************************
1320 tryRules :: SimplEnv -> [CoreRule]
1321 -> Id -> [OutExpr] -> SimplCont
1322 -> SimplM (Maybe (Arity, CoreExpr)) -- The arity is the number of
1323 -- args consumed by the rule
1324 tryRules env rules fn args call_cont
1328 = do { dflags <- getDOptsSmpl
1329 ; case activeRule dflags env of {
1330 Nothing -> return Nothing ; -- No rules apply
1332 case lookupRule act_fn (activeUnfInRule env) (getInScope env) fn args rules of {
1333 Nothing -> return Nothing ; -- No rule matches
1334 Just (rule, rule_rhs) ->
1336 do { tick (RuleFired (ru_name rule))
1337 ; trace_dump dflags rule rule_rhs $
1338 return (Just (ruleArity rule, rule_rhs)) }}}}
1340 trace_dump dflags rule rule_rhs stuff
1341 | not (dopt Opt_D_dump_rule_firings dflags) = stuff
1342 | not (dopt Opt_D_verbose_core2core dflags)
1344 = pprTrace "Rule fired:" (ftext (ru_name rule)) stuff
1346 = pprTrace "Rule fired"
1347 (vcat [text "Rule:" <+> ftext (ru_name rule),
1348 text "Before:" <+> ppr fn <+> sep (map pprParendExpr args),
1349 text "After: " <+> pprCoreExpr rule_rhs,
1350 text "Cont: " <+> ppr call_cont])
1354 Note [Rules for recursive functions]
1355 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1356 You might think that we shouldn't apply rules for a loop breaker:
1357 doing so might give rise to an infinite loop, because a RULE is
1358 rather like an extra equation for the function:
1359 RULE: f (g x) y = x+y
1362 But it's too drastic to disable rules for loop breakers.
1363 Even the foldr/build rule would be disabled, because foldr
1364 is recursive, and hence a loop breaker:
1365 foldr k z (build g) = g k z
1366 So it's up to the programmer: rules can cause divergence
1369 %************************************************************************
1371 Rebuilding a cse expression
1373 %************************************************************************
1375 Note [Case elimination]
1376 ~~~~~~~~~~~~~~~~~~~~~~~
1377 The case-elimination transformation discards redundant case expressions.
1378 Start with a simple situation:
1380 case x# of ===> e[x#/y#]
1383 (when x#, y# are of primitive type, of course). We can't (in general)
1384 do this for algebraic cases, because we might turn bottom into
1387 The code in SimplUtils.prepareAlts has the effect of generalise this
1388 idea to look for a case where we're scrutinising a variable, and we
1389 know that only the default case can match. For example:
1393 DEFAULT -> ...(case x of
1397 Here the inner case is first trimmed to have only one alternative, the
1398 DEFAULT, after which it's an instance of the previous case. This
1399 really only shows up in eliminating error-checking code.
1401 We also make sure that we deal with this very common case:
1406 Here we are using the case as a strict let; if x is used only once
1407 then we want to inline it. We have to be careful that this doesn't
1408 make the program terminate when it would have diverged before, so we
1410 - e is already evaluated (it may so if e is a variable)
1411 - x is used strictly, or
1413 Lastly, the code in SimplUtils.mkCase combines identical RHSs. So
1415 case e of ===> case e of DEFAULT -> r
1419 Now again the case may be elminated by the CaseElim transformation.
1422 Further notes about case elimination
1423 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1424 Consider: test :: Integer -> IO ()
1427 Turns out that this compiles to:
1430 eta1 :: State# RealWorld ->
1431 case PrelNum.< eta PrelNum.zeroInteger of wild { __DEFAULT ->
1433 (PrelNum.jtos eta ($w[] @ Char))
1435 of wild1 { (# new_s, a4 #) -> PrelIO.lvl23 new_s }}
1437 Notice the strange '<' which has no effect at all. This is a funny one.
1438 It started like this:
1440 f x y = if x < 0 then jtos x
1441 else if y==0 then "" else jtos x
1443 At a particular call site we have (f v 1). So we inline to get
1445 if v < 0 then jtos x
1446 else if 1==0 then "" else jtos x
1448 Now simplify the 1==0 conditional:
1450 if v<0 then jtos v else jtos v
1452 Now common-up the two branches of the case:
1454 case (v<0) of DEFAULT -> jtos v
1456 Why don't we drop the case? Because it's strict in v. It's technically
1457 wrong to drop even unnecessary evaluations, and in practice they
1458 may be a result of 'seq' so we *definitely* don't want to drop those.
1459 I don't really know how to improve this situation.
1462 ---------------------------------------------------------
1463 -- Eliminate the case if possible
1465 rebuildCase, reallyRebuildCase
1467 -> OutExpr -- Scrutinee
1468 -> InId -- Case binder
1469 -> [InAlt] -- Alternatives (inceasing order)
1471 -> SimplM (SimplEnv, OutExpr)
1473 --------------------------------------------------
1474 -- 1. Eliminate the case if there's a known constructor
1475 --------------------------------------------------
1477 rebuildCase env scrut case_bndr alts cont
1478 | Lit lit <- scrut -- No need for same treatment as constructors
1479 -- because literals are inlined more vigorously
1480 = do { tick (KnownBranch case_bndr)
1481 ; case findAlt (LitAlt lit) alts of
1482 Nothing -> missingAlt env case_bndr alts cont
1483 Just (_, bs, rhs) -> simple_rhs bs rhs }
1485 | Just (con, ty_args, other_args) <- exprIsConApp_maybe (activeUnfInRule env) scrut
1486 -- Works when the scrutinee is a variable with a known unfolding
1487 -- as well as when it's an explicit constructor application
1488 = do { tick (KnownBranch case_bndr)
1489 ; case findAlt (DataAlt con) alts of
1490 Nothing -> missingAlt env case_bndr alts cont
1491 Just (DEFAULT, bs, rhs) -> simple_rhs bs rhs
1492 Just (_, bs, rhs) -> knownCon env scrut con ty_args other_args
1493 case_bndr bs rhs cont
1496 simple_rhs bs rhs = ASSERT( null bs )
1497 do { env' <- simplNonRecX env case_bndr scrut
1498 ; simplExprF env' rhs cont }
1501 --------------------------------------------------
1502 -- 2. Eliminate the case if scrutinee is evaluated
1503 --------------------------------------------------
1505 rebuildCase env scrut case_bndr [(_, bndrs, rhs)] cont
1506 -- See if we can get rid of the case altogether
1507 -- See Note [Case elimination]
1508 -- mkCase made sure that if all the alternatives are equal,
1509 -- then there is now only one (DEFAULT) rhs
1510 | all isDeadBinder bndrs -- bndrs are [InId]
1512 -- Check that the scrutinee can be let-bound instead of case-bound
1513 , exprOkForSpeculation scrut
1514 -- OK not to evaluate it
1515 -- This includes things like (==# a# b#)::Bool
1516 -- so that we simplify
1517 -- case ==# a# b# of { True -> x; False -> x }
1520 -- This particular example shows up in default methods for
1521 -- comparision operations (e.g. in (>=) for Int.Int32)
1522 || exprIsHNF scrut -- It's already evaluated
1523 || var_demanded_later scrut -- It'll be demanded later
1525 -- || not opt_SimplPedanticBottoms) -- Or we don't care!
1526 -- We used to allow improving termination by discarding cases, unless -fpedantic-bottoms was on,
1527 -- but that breaks badly for the dataToTag# primop, which relies on a case to evaluate
1528 -- its argument: case x of { y -> dataToTag# y }
1529 -- Here we must *not* discard the case, because dataToTag# just fetches the tag from
1530 -- the info pointer. So we'll be pedantic all the time, and see if that gives any
1532 -- Also we don't want to discard 'seq's
1533 = do { tick (CaseElim case_bndr)
1534 ; env' <- simplNonRecX env case_bndr scrut
1535 ; simplExprF env' rhs cont }
1537 -- The case binder is going to be evaluated later,
1538 -- and the scrutinee is a simple variable
1539 var_demanded_later (Var v) = isStrictDmd (idDemandInfo case_bndr)
1540 && not (isTickBoxOp v)
1541 -- ugly hack; covering this case is what
1542 -- exprOkForSpeculation was intended for.
1543 var_demanded_later _ = False
1545 --------------------------------------------------
1546 -- 3. Try seq rules; see Note [User-defined RULES for seq] in MkId
1547 --------------------------------------------------
1549 rebuildCase env scrut case_bndr alts@[(_, bndrs, rhs)] cont
1550 | all isDeadBinder (case_bndr : bndrs) -- So this is just 'seq'
1551 = do { let rhs' = substExpr (text "rebuild-case") env rhs
1552 out_args = [Type (substTy env (idType case_bndr)),
1553 Type (exprType rhs'), scrut, rhs']
1554 -- Lazily evaluated, so we don't do most of this
1556 ; rule_base <- getSimplRules
1557 ; mb_rule <- tryRules env (getRules rule_base seqId) seqId out_args cont
1559 Just (n_args, res) -> simplExprF (zapSubstEnv env)
1560 (mkApps res (drop n_args out_args))
1562 Nothing -> reallyRebuildCase env scrut case_bndr alts cont }
1564 rebuildCase env scrut case_bndr alts cont
1565 = reallyRebuildCase env scrut case_bndr alts cont
1567 --------------------------------------------------
1568 -- 3. Catch-all case
1569 --------------------------------------------------
1571 reallyRebuildCase env scrut case_bndr alts cont
1572 = do { -- Prepare the continuation;
1573 -- The new subst_env is in place
1574 (env', dup_cont, nodup_cont) <- prepareCaseCont env alts cont
1576 -- Simplify the alternatives
1577 ; (scrut', case_bndr', alts') <- simplAlts env' scrut case_bndr alts dup_cont
1579 -- Check for empty alternatives
1580 ; if null alts' then missingAlt env case_bndr alts cont
1582 { dflags <- getDOptsSmpl
1583 ; case_expr <- mkCase dflags scrut' case_bndr' alts'
1585 -- Notice that rebuild gets the in-scope set from env', not alt_env
1586 -- (which in any case is only build in simplAlts)
1587 -- The case binder *not* scope over the whole returned case-expression
1588 ; rebuild env' case_expr nodup_cont } }
1591 simplCaseBinder checks whether the scrutinee is a variable, v. If so,
1592 try to eliminate uses of v in the RHSs in favour of case_bndr; that
1593 way, there's a chance that v will now only be used once, and hence
1596 Historical note: we use to do the "case binder swap" in the Simplifier
1597 so there were additional complications if the scrutinee was a variable.
1598 Now the binder-swap stuff is done in the occurrence analyer; see
1599 OccurAnal Note [Binder swap].
1603 If the case binder is not dead, then neither are the pattern bound
1605 case <any> of x { (a,b) ->
1606 case x of { (p,q) -> p } }
1607 Here (a,b) both look dead, but come alive after the inner case is eliminated.
1608 The point is that we bring into the envt a binding
1610 after the outer case, and that makes (a,b) alive. At least we do unless
1611 the case binder is guaranteed dead.
1613 In practice, the scrutinee is almost always a variable, so we pretty
1614 much always zap the OccInfo of the binders. It doesn't matter much though.
1619 Consider case (v `cast` co) of x { I# y ->
1620 ... (case (v `cast` co) of {...}) ...
1621 We'd like to eliminate the inner case. We can get this neatly by
1622 arranging that inside the outer case we add the unfolding
1623 v |-> x `cast` (sym co)
1624 to v. Then we should inline v at the inner case, cancel the casts, and away we go
1626 Note [Improving seq]
1629 type family F :: * -> *
1630 type instance F Int = Int
1632 ... case e of x { DEFAULT -> rhs } ...
1634 where x::F Int. Then we'd like to rewrite (F Int) to Int, getting
1636 case e `cast` co of x'::Int
1637 I# x# -> let x = x' `cast` sym co
1640 so that 'rhs' can take advantage of the form of x'.
1642 Notice that Note [Case of cast] may then apply to the result.
1644 Nota Bene: We only do the [Improving seq] transformation if the
1645 case binder 'x' is actually used in the rhs; that is, if the case
1646 is *not* a *pure* seq.
1647 a) There is no point in adding the cast to a pure seq.
1648 b) There is a good reason not to: doing so would interfere
1649 with seq rules (Note [Built-in RULES for seq] in MkId).
1650 In particular, this [Improving seq] thing *adds* a cast
1651 while [Built-in RULES for seq] *removes* one, so they
1654 You might worry about
1655 case v of x { __DEFAULT ->
1656 ... case (v `cast` co) of y { I# -> ... }}
1657 This is a pure seq (since x is unused), so [Improving seq] won't happen.
1658 But it's ok: the simplifier will replace 'v' by 'x' in the rhs to get
1659 case v of x { __DEFAULT ->
1660 ... case (x `cast` co) of y { I# -> ... }}
1661 Now the outer case is not a pure seq, so [Improving seq] will happen,
1662 and then the inner case will disappear.
1664 The need for [Improving seq] showed up in Roman's experiments. Example:
1665 foo :: F Int -> Int -> Int
1666 foo t n = t `seq` bar n
1669 bar n = bar (n - case t of TI i -> i)
1670 Here we'd like to avoid repeated evaluating t inside the loop, by
1671 taking advantage of the `seq`.
1673 At one point I did transformation in LiberateCase, but it's more
1674 robust here. (Otherwise, there's a danger that we'll simply drop the
1675 'seq' altogether, before LiberateCase gets to see it.)
1678 simplAlts :: SimplEnv
1680 -> InId -- Case binder
1681 -> [InAlt] -- Non-empty
1683 -> SimplM (OutExpr, OutId, [OutAlt]) -- Includes the continuation
1684 -- Like simplExpr, this just returns the simplified alternatives;
1685 -- it does not return an environment
1687 simplAlts env scrut case_bndr alts cont'
1688 = -- pprTrace "simplAlts" (ppr alts $$ ppr (seTvSubst env)) $
1689 do { let env0 = zapFloats env
1691 ; (env1, case_bndr1) <- simplBinder env0 case_bndr
1693 ; fam_envs <- getFamEnvs
1694 ; (alt_env', scrut', case_bndr') <- improveSeq fam_envs env1 scrut
1695 case_bndr case_bndr1 alts
1697 ; (imposs_deflt_cons, in_alts) <- prepareAlts scrut' case_bndr' alts
1699 ; alts' <- mapM (simplAlt alt_env' imposs_deflt_cons case_bndr' cont') in_alts
1700 ; return (scrut', case_bndr', alts') }
1703 ------------------------------------
1704 improveSeq :: (FamInstEnv, FamInstEnv) -> SimplEnv
1705 -> OutExpr -> InId -> OutId -> [InAlt]
1706 -> SimplM (SimplEnv, OutExpr, OutId)
1707 -- Note [Improving seq]
1708 improveSeq fam_envs env scrut case_bndr case_bndr1 [(DEFAULT,_,_)]
1709 | not (isDeadBinder case_bndr) -- Not a pure seq! See the Note!
1710 , Just (co, ty2) <- topNormaliseType fam_envs (idType case_bndr1)
1711 = do { case_bndr2 <- newId (fsLit "nt") ty2
1712 ; let rhs = DoneEx (Var case_bndr2 `Cast` mkSymCoercion co)
1713 env2 = extendIdSubst env case_bndr rhs
1714 ; return (env2, scrut `Cast` co, case_bndr2) }
1716 improveSeq _ env scrut _ case_bndr1 _
1717 = return (env, scrut, case_bndr1)
1720 ------------------------------------
1721 simplAlt :: SimplEnv
1722 -> [AltCon] -- These constructors can't be present when
1723 -- matching the DEFAULT alternative
1724 -> OutId -- The case binder
1729 simplAlt env imposs_deflt_cons case_bndr' cont' (DEFAULT, bndrs, rhs)
1730 = ASSERT( null bndrs )
1731 do { let env' = addBinderOtherCon env case_bndr' imposs_deflt_cons
1732 -- Record the constructors that the case-binder *can't* be.
1733 ; rhs' <- simplExprC env' rhs cont'
1734 ; return (DEFAULT, [], rhs') }
1736 simplAlt env _ case_bndr' cont' (LitAlt lit, bndrs, rhs)
1737 = ASSERT( null bndrs )
1738 do { let env' = addBinderUnfolding env case_bndr' (Lit lit)
1739 ; rhs' <- simplExprC env' rhs cont'
1740 ; return (LitAlt lit, [], rhs') }
1742 simplAlt env _ case_bndr' cont' (DataAlt con, vs, rhs)
1743 = do { -- Deal with the pattern-bound variables
1744 -- Mark the ones that are in ! positions in the
1745 -- data constructor as certainly-evaluated.
1746 -- NB: simplLamBinders preserves this eval info
1747 let vs_with_evals = add_evals (dataConRepStrictness con)
1748 ; (env', vs') <- simplLamBndrs env vs_with_evals
1750 -- Bind the case-binder to (con args)
1751 ; let inst_tys' = tyConAppArgs (idType case_bndr')
1752 con_args = map Type inst_tys' ++ varsToCoreExprs vs'
1753 env'' = addBinderUnfolding env' case_bndr'
1754 (mkConApp con con_args)
1756 ; rhs' <- simplExprC env'' rhs cont'
1757 ; return (DataAlt con, vs', rhs') }
1759 -- add_evals records the evaluated-ness of the bound variables of
1760 -- a case pattern. This is *important*. Consider
1761 -- data T = T !Int !Int
1763 -- case x of { T a b -> T (a+1) b }
1765 -- We really must record that b is already evaluated so that we don't
1766 -- go and re-evaluate it when constructing the result.
1767 -- See Note [Data-con worker strictness] in MkId.lhs
1772 go (v:vs') strs | isTyCoVar v = v : go vs' strs
1773 go (v:vs') (str:strs)
1774 | isMarkedStrict str = evald_v : go vs' strs
1775 | otherwise = zapped_v : go vs' strs
1777 zapped_v = zap_occ_info v
1778 evald_v = zapped_v `setIdUnfolding` evaldUnfolding
1779 go _ _ = pprPanic "cat_evals" (ppr con $$ ppr vs $$ ppr the_strs)
1781 -- See Note [zapOccInfo]
1782 -- zap_occ_info: if the case binder is alive, then we add the unfolding
1784 -- to the envt; so vs are now very much alive
1785 -- Note [Aug06] I can't see why this actually matters, but it's neater
1786 -- case e of t { (a,b) -> ...(case t of (p,q) -> p)... }
1787 -- ==> case e of t { (a,b) -> ...(a)... }
1788 -- Look, Ma, a is alive now.
1789 zap_occ_info = zapCasePatIdOcc case_bndr'
1791 addBinderUnfolding :: SimplEnv -> Id -> CoreExpr -> SimplEnv
1792 addBinderUnfolding env bndr rhs
1793 = modifyInScope env (bndr `setIdUnfolding` mkSimpleUnfolding rhs)
1795 addBinderOtherCon :: SimplEnv -> Id -> [AltCon] -> SimplEnv
1796 addBinderOtherCon env bndr cons
1797 = modifyInScope env (bndr `setIdUnfolding` mkOtherCon cons)
1799 zapCasePatIdOcc :: Id -> Id -> Id
1800 -- Consider case e of b { (a,b) -> ... }
1801 -- Then if we bind b to (a,b) in "...", and b is not dead,
1802 -- then we must zap the deadness info on a,b
1803 zapCasePatIdOcc case_bndr
1804 | isDeadBinder case_bndr = \ pat_id -> pat_id
1805 | otherwise = \ pat_id -> zapIdOccInfo pat_id
1809 %************************************************************************
1811 \subsection{Known constructor}
1813 %************************************************************************
1815 We are a bit careful with occurrence info. Here's an example
1817 (\x* -> case x of (a*, b) -> f a) (h v, e)
1819 where the * means "occurs once". This effectively becomes
1820 case (h v, e) of (a*, b) -> f a)
1822 let a* = h v; b = e in f a
1826 All this should happen in one sweep.
1829 knownCon :: SimplEnv
1830 -> OutExpr -- The scrutinee
1831 -> DataCon -> [OutType] -> [OutExpr] -- The scrutinee (in pieces)
1832 -> InId -> [InBndr] -> InExpr -- The alternative
1834 -> SimplM (SimplEnv, OutExpr)
1836 knownCon env scrut dc dc_ty_args dc_args bndr bs rhs cont
1837 = do { env' <- bind_args env bs dc_args
1838 ; env'' <- bind_case_bndr env'
1839 ; simplExprF env'' rhs cont }
1841 zap_occ = zapCasePatIdOcc bndr -- bndr is an InId
1844 bind_args env' [] _ = return env'
1846 bind_args env' (b:bs') (Type ty : args)
1847 = ASSERT( isTyCoVar b )
1848 bind_args (extendTvSubst env' b ty) bs' args
1850 bind_args env' (b:bs') (arg : args)
1852 do { let b' = zap_occ b
1853 -- Note that the binder might be "dead", because it doesn't
1854 -- occur in the RHS; and simplNonRecX may therefore discard
1855 -- it via postInlineUnconditionally.
1856 -- Nevertheless we must keep it if the case-binder is alive,
1857 -- because it may be used in the con_app. See Note [zapOccInfo]
1858 ; env'' <- simplNonRecX env' b' arg
1859 ; bind_args env'' bs' args }
1862 pprPanic "bind_args" $ ppr dc $$ ppr bs $$ ppr dc_args $$
1863 text "scrut:" <+> ppr scrut
1865 -- It's useful to bind bndr to scrut, rather than to a fresh
1866 -- binding x = Con arg1 .. argn
1867 -- because very often the scrut is a variable, so we avoid
1868 -- creating, and then subsequently eliminating, a let-binding
1869 -- BUT, if scrut is a not a variable, we must be careful
1870 -- about duplicating the arg redexes; in that case, make
1871 -- a new con-app from the args
1873 | isDeadBinder bndr = return env
1874 | exprIsTrivial scrut = return (extendIdSubst env bndr (DoneEx scrut))
1875 | otherwise = do { dc_args <- mapM (simplVar env) bs
1876 -- dc_ty_args are aready OutTypes,
1877 -- but bs are InBndrs
1878 ; let con_app = Var (dataConWorkId dc)
1879 `mkTyApps` dc_ty_args
1881 ; simplNonRecX env bndr con_app }
1884 missingAlt :: SimplEnv -> Id -> [InAlt] -> SimplCont -> SimplM (SimplEnv, OutExpr)
1885 -- This isn't strictly an error, although it is unusual.
1886 -- It's possible that the simplifer might "see" that
1887 -- an inner case has no accessible alternatives before
1888 -- it "sees" that the entire branch of an outer case is
1889 -- inaccessible. So we simply put an error case here instead.
1890 missingAlt env case_bndr alts cont
1891 = WARN( True, ptext (sLit "missingAlt") <+> ppr case_bndr )
1892 return (env, mkImpossibleExpr res_ty)
1894 res_ty = contResultType env (substTy env (coreAltsType alts)) cont
1898 %************************************************************************
1900 \subsection{Duplicating continuations}
1902 %************************************************************************
1905 prepareCaseCont :: SimplEnv
1906 -> [InAlt] -> SimplCont
1907 -> SimplM (SimplEnv, SimplCont,SimplCont)
1908 -- Return a duplicatable continuation, a non-duplicable part
1909 -- plus some extra bindings (that scope over the entire
1912 -- No need to make it duplicatable if there's only one alternative
1913 prepareCaseCont env [_] cont = return (env, cont, mkBoringStop)
1914 prepareCaseCont env _ cont = mkDupableCont env cont
1918 mkDupableCont :: SimplEnv -> SimplCont
1919 -> SimplM (SimplEnv, SimplCont, SimplCont)
1921 mkDupableCont env cont
1922 | contIsDupable cont
1923 = return (env, cont, mkBoringStop)
1925 mkDupableCont _ (Stop {}) = panic "mkDupableCont" -- Handled by previous eqn
1927 mkDupableCont env (CoerceIt ty cont)
1928 = do { (env', dup, nodup) <- mkDupableCont env cont
1929 ; return (env', CoerceIt ty dup, nodup) }
1931 mkDupableCont env cont@(StrictBind {})
1932 = return (env, mkBoringStop, cont)
1933 -- See Note [Duplicating StrictBind]
1935 mkDupableCont env (StrictArg info cci cont)
1936 -- See Note [Duplicating StrictArg]
1937 = do { (env', dup, nodup) <- mkDupableCont env cont
1938 ; (env'', args') <- mapAccumLM (makeTrivial NotTopLevel) env' (ai_args info)
1939 ; return (env'', StrictArg (info { ai_args = args' }) cci dup, nodup) }
1941 mkDupableCont env (ApplyTo _ arg se cont)
1942 = -- e.g. [...hole...] (...arg...)
1944 -- let a = ...arg...
1945 -- in [...hole...] a
1946 do { (env', dup_cont, nodup_cont) <- mkDupableCont env cont
1947 ; arg' <- simplExpr (se `setInScope` env') arg
1948 ; (env'', arg'') <- makeTrivial NotTopLevel env' arg'
1949 ; let app_cont = ApplyTo OkToDup arg'' (zapSubstEnv env'') dup_cont
1950 ; return (env'', app_cont, nodup_cont) }
1952 mkDupableCont env cont@(Select _ case_bndr [(_, bs, _rhs)] _ _)
1953 -- See Note [Single-alternative case]
1954 -- | not (exprIsDupable rhs && contIsDupable case_cont)
1955 -- | not (isDeadBinder case_bndr)
1956 | all isDeadBinder bs -- InIds
1957 && not (isUnLiftedType (idType case_bndr))
1958 -- Note [Single-alternative-unlifted]
1959 = return (env, mkBoringStop, cont)
1961 mkDupableCont env (Select _ case_bndr alts se cont)
1962 = -- e.g. (case [...hole...] of { pi -> ei })
1964 -- let ji = \xij -> ei
1965 -- in case [...hole...] of { pi -> ji xij }
1966 do { tick (CaseOfCase case_bndr)
1967 ; (env', dup_cont, nodup_cont) <- mkDupableCont env cont
1968 -- NB: call mkDupableCont here, *not* prepareCaseCont
1969 -- We must make a duplicable continuation, whereas prepareCaseCont
1970 -- doesn't when there is a single case branch
1972 ; let alt_env = se `setInScope` env'
1973 ; (alt_env', case_bndr') <- simplBinder alt_env case_bndr
1974 ; alts' <- mapM (simplAlt alt_env' [] case_bndr' dup_cont) alts
1975 -- Safe to say that there are no handled-cons for the DEFAULT case
1976 -- NB: simplBinder does not zap deadness occ-info, so
1977 -- a dead case_bndr' will still advertise its deadness
1978 -- This is really important because in
1979 -- case e of b { (# p,q #) -> ... }
1980 -- b is always dead, and indeed we are not allowed to bind b to (# p,q #),
1981 -- which might happen if e was an explicit unboxed pair and b wasn't marked dead.
1982 -- In the new alts we build, we have the new case binder, so it must retain
1984 -- NB: we don't use alt_env further; it has the substEnv for
1985 -- the alternatives, and we don't want that
1987 ; (env'', alts'') <- mkDupableAlts env' case_bndr' alts'
1988 ; return (env'', -- Note [Duplicated env]
1989 Select OkToDup case_bndr' alts'' (zapSubstEnv env'') mkBoringStop,
1993 mkDupableAlts :: SimplEnv -> OutId -> [InAlt]
1994 -> SimplM (SimplEnv, [InAlt])
1995 -- Absorbs the continuation into the new alternatives
1997 mkDupableAlts env case_bndr' the_alts
2000 go env0 [] = return (env0, [])
2002 = do { (env1, alt') <- mkDupableAlt env0 case_bndr' alt
2003 ; (env2, alts') <- go env1 alts
2004 ; return (env2, alt' : alts' ) }
2006 mkDupableAlt :: SimplEnv -> OutId -> (AltCon, [CoreBndr], CoreExpr)
2007 -> SimplM (SimplEnv, (AltCon, [CoreBndr], CoreExpr))
2008 mkDupableAlt env case_bndr (con, bndrs', rhs')
2009 | exprIsDupable rhs' -- Note [Small alternative rhs]
2010 = return (env, (con, bndrs', rhs'))
2012 = do { let rhs_ty' = exprType rhs'
2013 scrut_ty = idType case_bndr
2016 DEFAULT -> case_bndr
2017 DataAlt dc -> setIdUnfolding case_bndr unf
2019 -- See Note [Case binders and join points]
2020 unf = mkInlineUnfolding Nothing rhs
2021 rhs = mkConApp dc (map Type (tyConAppArgs scrut_ty)
2022 ++ varsToCoreExprs bndrs')
2024 LitAlt {} -> WARN( True, ptext (sLit "mkDupableAlt")
2025 <+> ppr case_bndr <+> ppr con )
2027 -- The case binder is alive but trivial, so why has
2028 -- it not been substituted away?
2030 used_bndrs' | isDeadBinder case_bndr = filter abstract_over bndrs'
2031 | otherwise = bndrs' ++ [case_bndr_w_unf]
2034 | isTyCoVar bndr = True -- Abstract over all type variables just in case
2035 | otherwise = not (isDeadBinder bndr)
2036 -- The deadness info on the new Ids is preserved by simplBinders
2038 ; (final_bndrs', final_args) -- Note [Join point abstraction]
2039 <- if (any isId used_bndrs')
2040 then return (used_bndrs', varsToCoreExprs used_bndrs')
2041 else do { rw_id <- newId (fsLit "w") realWorldStatePrimTy
2042 ; return ([rw_id], [Var realWorldPrimId]) }
2044 ; join_bndr <- newId (fsLit "$j") (mkPiTypes final_bndrs' rhs_ty')
2045 -- Note [Funky mkPiTypes]
2047 ; let -- We make the lambdas into one-shot-lambdas. The
2048 -- join point is sure to be applied at most once, and doing so
2049 -- prevents the body of the join point being floated out by
2050 -- the full laziness pass
2051 really_final_bndrs = map one_shot final_bndrs'
2052 one_shot v | isId v = setOneShotLambda v
2054 join_rhs = mkLams really_final_bndrs rhs'
2055 join_call = mkApps (Var join_bndr) final_args
2057 ; env' <- addPolyBind NotTopLevel env (NonRec join_bndr join_rhs)
2058 ; return (env', (con, bndrs', join_call)) }
2059 -- See Note [Duplicated env]
2062 Note [Case binders and join points]
2063 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2065 case (case .. ) of c {
2068 If we make a join point with c but not c# we get
2069 $j = \c -> ....c....
2071 But if later inlining scrutines the c, thus
2073 $j = \c -> ... case c of { I# y -> ... } ...
2075 we won't see that 'c' has already been scrutinised. This actually
2076 happens in the 'tabulate' function in wave4main, and makes a significant
2077 difference to allocation.
2079 An alternative plan is this:
2081 $j = \c# -> let c = I# c# in ...c....
2083 but that is bad if 'c' is *not* later scrutinised.
2085 So instead we do both: we pass 'c' and 'c#' , and record in c's inlining
2086 (an InlineRule) that it's really I# c#, thus
2088 $j = \c# -> \c[=I# c#] -> ...c....
2090 Absence analysis may later discard 'c'.
2092 NB: take great care when doing strictness analysis;
2093 see Note [Lamba-bound unfoldings] in DmdAnal.
2095 Also note that we can still end up passing stuff that isn't used. Before
2096 strictness analysis we have
2097 let $j x y c{=(x,y)} = (h c, ...)
2099 After strictness analysis we see that h is strict, we end up with
2100 let $j x y c{=(x,y)} = ($wh x y, ...)
2103 Note [Duplicated env]
2104 ~~~~~~~~~~~~~~~~~~~~~
2105 Some of the alternatives are simplified, but have not been turned into a join point
2106 So they *must* have an zapped subst-env. So we can't use completeNonRecX to
2107 bind the join point, because it might to do PostInlineUnconditionally, and
2108 we'd lose that when zapping the subst-env. We could have a per-alt subst-env,
2109 but zapping it (as we do in mkDupableCont, the Select case) is safe, and
2110 at worst delays the join-point inlining.
2112 Note [Small alternative rhs]
2113 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2114 It is worth checking for a small RHS because otherwise we
2115 get extra let bindings that may cause an extra iteration of the simplifier to
2116 inline back in place. Quite often the rhs is just a variable or constructor.
2117 The Ord instance of Maybe in PrelMaybe.lhs, for example, took several extra
2118 iterations because the version with the let bindings looked big, and so wasn't
2119 inlined, but after the join points had been inlined it looked smaller, and so
2122 NB: we have to check the size of rhs', not rhs.
2123 Duplicating a small InAlt might invalidate occurrence information
2124 However, if it *is* dupable, we return the *un* simplified alternative,
2125 because otherwise we'd need to pair it up with an empty subst-env....
2126 but we only have one env shared between all the alts.
2127 (Remember we must zap the subst-env before re-simplifying something).
2128 Rather than do this we simply agree to re-simplify the original (small) thing later.
2130 Note [Funky mkPiTypes]
2131 ~~~~~~~~~~~~~~~~~~~~~~
2132 Notice the funky mkPiTypes. If the contructor has existentials
2133 it's possible that the join point will be abstracted over
2134 type varaibles as well as term variables.
2135 Example: Suppose we have
2136 data T = forall t. C [t]
2138 case (case e of ...) of
2140 We get the join point
2141 let j :: forall t. [t] -> ...
2142 j = /\t \xs::[t] -> rhs
2144 case (case e of ...) of
2145 C t xs::[t] -> j t xs
2147 Note [Join point abstaction]
2148 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2149 If we try to lift a primitive-typed something out
2150 for let-binding-purposes, we will *caseify* it (!),
2151 with potentially-disastrous strictness results. So
2152 instead we turn it into a function: \v -> e
2153 where v::State# RealWorld#. The value passed to this function
2154 is realworld#, which generates (almost) no code.
2156 There's a slight infelicity here: we pass the overall
2157 case_bndr to all the join points if it's used in *any* RHS,
2158 because we don't know its usage in each RHS separately
2160 We used to say "&& isUnLiftedType rhs_ty'" here, but now
2161 we make the join point into a function whenever used_bndrs'
2162 is empty. This makes the join-point more CPR friendly.
2163 Consider: let j = if .. then I# 3 else I# 4
2164 in case .. of { A -> j; B -> j; C -> ... }
2166 Now CPR doesn't w/w j because it's a thunk, so
2167 that means that the enclosing function can't w/w either,
2168 which is a lose. Here's the example that happened in practice:
2169 kgmod :: Int -> Int -> Int
2170 kgmod x y = if x > 0 && y < 0 || x < 0 && y > 0
2174 I have seen a case alternative like this:
2176 It's a bit silly to add the realWorld dummy arg in this case, making
2179 (the \v alone is enough to make CPR happy) but I think it's rare
2181 Note [Duplicating StrictArg]
2182 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2183 The original plan had (where E is a big argument)
2185 ==> let $j = \a -> f E a
2188 But this is terrible! Here's an example:
2189 && E (case x of { T -> F; F -> T })
2190 Now, && is strict so we end up simplifying the case with
2191 an ArgOf continuation. If we let-bind it, we get
2192 let $j = \v -> && E v
2193 in simplExpr (case x of { T -> F; F -> T })
2195 And after simplifying more we get
2196 let $j = \v -> && E v
2197 in case x of { T -> $j F; F -> $j T }
2198 Which is a Very Bad Thing
2200 What we do now is this
2204 Now if the thing in the hole is a case expression (which is when
2205 we'll call mkDupableCont), we'll push the function call into the
2206 branches, which is what we want. Now RULES for f may fire, and
2207 call-pattern specialisation. Here's an example from Trac #3116
2210 _ -> Chunk p fpc (o+1) (l-1) bs')
2211 If we can push the call for 'go' inside the case, we get
2212 call-pattern specialisation for 'go', which is *crucial* for
2215 Here is the (&&) example:
2216 && E (case x of { T -> F; F -> T })
2218 case x of { T -> && a F; F -> && a T }
2222 * Arguments to f *after* the strict one are handled by
2223 the ApplyTo case of mkDupableCont. Eg
2226 * We can only do the let-binding of E because the function
2227 part of a StrictArg continuation is an explicit syntax
2228 tree. In earlier versions we represented it as a function
2229 (CoreExpr -> CoreEpxr) which we couldn't take apart.
2231 Do *not* duplicate StrictBind and StritArg continuations. We gain
2232 nothing by propagating them into the expressions, and we do lose a
2235 The desire not to duplicate is the entire reason that
2236 mkDupableCont returns a pair of continuations.
2238 Note [Duplicating StrictBind]
2239 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2240 Unlike StrictArg, there doesn't seem anything to gain from
2241 duplicating a StrictBind continuation, so we don't.
2243 The desire not to duplicate is the entire reason that
2244 mkDupableCont returns a pair of continuations.
2247 Note [Single-alternative cases]
2248 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2249 This case is just like the ArgOf case. Here's an example:
2253 case (case x of I# x' ->
2255 True -> I# (negate# x')
2256 False -> I# x') of y {
2258 Because the (case x) has only one alternative, we'll transform to
2260 case (case x' <# 0# of
2261 True -> I# (negate# x')
2262 False -> I# x') of y {
2264 But now we do *NOT* want to make a join point etc, giving
2266 let $j = \y -> MkT y
2268 True -> $j (I# (negate# x'))
2270 In this case the $j will inline again, but suppose there was a big
2271 strict computation enclosing the orginal call to MkT. Then, it won't
2272 "see" the MkT any more, because it's big and won't get duplicated.
2273 And, what is worse, nothing was gained by the case-of-case transform.
2275 So, in circumstances like these, we don't want to build join points
2276 and push the outer case into the branches of the inner one. Instead,
2277 don't duplicate the continuation.
2279 When should we use this strategy? We should not use it on *every*
2280 single-alternative case:
2281 e.g. case (case ....) of (a,b) -> (# a,b #)
2282 Here we must push the outer case into the inner one!
2285 * Match [(DEFAULT,_,_)], but in the common case of Int,
2286 the alternative-filling-in code turned the outer case into
2287 case (...) of y { I# _ -> MkT y }
2289 * Match on single alternative plus (not (isDeadBinder case_bndr))
2290 Rationale: pushing the case inwards won't eliminate the construction.
2291 But there's a risk of
2292 case (...) of y { (a,b) -> let z=(a,b) in ... }
2293 Now y looks dead, but it'll come alive again. Still, this
2294 seems like the best option at the moment.
2296 * Match on single alternative plus (all (isDeadBinder bndrs))
2297 Rationale: this is essentially seq.
2299 * Match when the rhs is *not* duplicable, and hence would lead to a
2300 join point. This catches the disaster-case above. We can test
2301 the *un-simplified* rhs, which is fine. It might get bigger or
2302 smaller after simplification; if it gets smaller, this case might
2303 fire next time round. NB also that we must test contIsDupable
2304 case_cont *too, because case_cont might be big!
2306 HOWEVER: I found that this version doesn't work well, because
2307 we can get let x = case (...) of { small } in ...case x...
2308 When x is inlined into its full context, we find that it was a bad
2309 idea to have pushed the outer case inside the (...) case.
2311 Note [Single-alternative-unlifted]
2312 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2313 Here's another single-alternative where we really want to do case-of-case:
2321 case y_s6X of tpl_s7m {
2322 M1.Mk1 ipv_s70 -> ipv_s70;
2323 M1.Mk2 ipv_s72 -> ipv_s72;
2329 case x_s74 of tpl_s7n {
2330 M1.Mk1 ipv_s77 -> ipv_s77;
2331 M1.Mk2 ipv_s79 -> ipv_s79;
2335 { __DEFAULT -> ==# [wild1_s7b wild_s7c];
2339 So the outer case is doing *nothing at all*, other than serving as a
2340 join-point. In this case we really want to do case-of-case and decide
2341 whether to use a real join point or just duplicate the continuation.
2343 Hence: check whether the case binder's type is unlifted, because then
2344 the outer case is *not* a seq.