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 )
16 import FamInstEnv ( FamInstEnv )
18 import MkId ( mkImpossibleExpr, seqId )
21 import Name ( mkSystemVarName )
23 import FamInstEnv ( topNormaliseType )
24 import DataCon ( DataCon, dataConWorkId, dataConRepStrictness )
26 import Demand ( isStrictDmd, splitStrictSig )
27 import PprCore ( pprParendExpr, pprCoreExpr )
28 import CoreUnfold ( mkUnfolding, mkCoreUnfolding, mkInlineRule,
29 exprIsConApp_maybe, callSiteInline, CallCtxt(..) )
31 import qualified CoreSubst
32 import CoreArity ( exprArity )
33 import Rules ( lookupRule, getRules )
34 import BasicTypes ( isMarkedStrict, Arity )
35 import CostCentre ( currentCCS, pushCCisNop )
36 import TysPrim ( realWorldStatePrimTy )
37 import PrelInfo ( realWorldPrimId )
38 import BasicTypes ( TopLevelFlag(..), isTopLevel, RecFlag(..) )
39 import MonadUtils ( foldlM, mapAccumLM )
40 import Maybes ( orElse )
41 import Data.List ( mapAccumL )
47 The guts of the simplifier is in this module, but the driver loop for
48 the simplifier is in SimplCore.lhs.
51 -----------------------------------------
52 *** IMPORTANT NOTE ***
53 -----------------------------------------
54 The simplifier used to guarantee that the output had no shadowing, but
55 it does not do so any more. (Actually, it never did!) The reason is
56 documented with simplifyArgs.
59 -----------------------------------------
60 *** IMPORTANT NOTE ***
61 -----------------------------------------
62 Many parts of the simplifier return a bunch of "floats" as well as an
63 expression. This is wrapped as a datatype SimplUtils.FloatsWith.
65 All "floats" are let-binds, not case-binds, but some non-rec lets may
66 be unlifted (with RHS ok-for-speculation).
70 -----------------------------------------
71 ORGANISATION OF FUNCTIONS
72 -----------------------------------------
74 - simplify all top-level binders
75 - for NonRec, call simplRecOrTopPair
76 - for Rec, call simplRecBind
79 ------------------------------
80 simplExpr (applied lambda) ==> simplNonRecBind
81 simplExpr (Let (NonRec ...) ..) ==> simplNonRecBind
82 simplExpr (Let (Rec ...) ..) ==> simplify binders; simplRecBind
84 ------------------------------
85 simplRecBind [binders already simplfied]
86 - use simplRecOrTopPair on each pair in turn
88 simplRecOrTopPair [binder already simplified]
89 Used for: recursive bindings (top level and nested)
90 top-level non-recursive bindings
92 - check for PreInlineUnconditionally
96 Used for: non-top-level non-recursive bindings
97 beta reductions (which amount to the same thing)
98 Because it can deal with strict arts, it takes a
99 "thing-inside" and returns an expression
101 - check for PreInlineUnconditionally
102 - simplify binder, including its IdInfo
111 simplNonRecX: [given a *simplified* RHS, but an *unsimplified* binder]
112 Used for: binding case-binder and constr args in a known-constructor case
113 - check for PreInLineUnconditionally
117 ------------------------------
118 simplLazyBind: [binder already simplified, RHS not]
119 Used for: recursive bindings (top level and nested)
120 top-level non-recursive bindings
121 non-top-level, but *lazy* non-recursive bindings
122 [must not be strict or unboxed]
123 Returns floats + an augmented environment, not an expression
124 - substituteIdInfo and add result to in-scope
125 [so that rules are available in rec rhs]
128 - float if exposes constructor or PAP
132 completeNonRecX: [binder and rhs both simplified]
133 - if the the thing needs case binding (unlifted and not ok-for-spec)
139 completeBind: [given a simplified RHS]
140 [used for both rec and non-rec bindings, top level and not]
141 - try PostInlineUnconditionally
142 - add unfolding [this is the only place we add an unfolding]
147 Right hand sides and arguments
148 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
149 In many ways we want to treat
150 (a) the right hand side of a let(rec), and
151 (b) a function argument
152 in the same way. But not always! In particular, we would
153 like to leave these arguments exactly as they are, so they
154 will match a RULE more easily.
159 It's harder to make the rule match if we ANF-ise the constructor,
160 or eta-expand the PAP:
162 f (let { a = g x; b = h x } in (a,b))
165 On the other hand if we see the let-defns
170 then we *do* want to ANF-ise and eta-expand, so that p and q
171 can be safely inlined.
173 Even floating lets out is a bit dubious. For let RHS's we float lets
174 out if that exposes a value, so that the value can be inlined more vigorously.
177 r = let x = e in (x,x)
179 Here, if we float the let out we'll expose a nice constructor. We did experiments
180 that showed this to be a generally good thing. But it was a bad thing to float
181 lets out unconditionally, because that meant they got allocated more often.
183 For function arguments, there's less reason to expose a constructor (it won't
184 get inlined). Just possibly it might make a rule match, but I'm pretty skeptical.
185 So for the moment we don't float lets out of function arguments either.
190 For eta expansion, we want to catch things like
192 case e of (a,b) -> \x -> case a of (p,q) -> \y -> r
194 If the \x was on the RHS of a let, we'd eta expand to bring the two
195 lambdas together. And in general that's a good thing to do. Perhaps
196 we should eta expand wherever we find a (value) lambda? Then the eta
197 expansion at a let RHS can concentrate solely on the PAP case.
200 %************************************************************************
202 \subsection{Bindings}
204 %************************************************************************
207 simplTopBinds :: SimplEnv -> [InBind] -> SimplM SimplEnv
209 simplTopBinds env0 binds0
210 = do { -- Put all the top-level binders into scope at the start
211 -- so that if a transformation rule has unexpectedly brought
212 -- anything into scope, then we don't get a complaint about that.
213 -- It's rather as if the top-level binders were imported.
214 ; env1 <- simplRecBndrs env0 (bindersOfBinds binds0)
215 ; dflags <- getDOptsSmpl
216 ; let dump_flag = dopt Opt_D_dump_inlinings dflags ||
217 dopt Opt_D_dump_rule_firings dflags
218 ; env2 <- simpl_binds dump_flag env1 binds0
219 ; freeTick SimplifierDone
222 -- We need to track the zapped top-level binders, because
223 -- they should have their fragile IdInfo zapped (notably occurrence info)
224 -- That's why we run down binds and bndrs' simultaneously.
226 -- The dump-flag emits a trace for each top-level binding, which
227 -- helps to locate the tracing for inlining and rule firing
228 simpl_binds :: Bool -> SimplEnv -> [InBind] -> SimplM SimplEnv
229 simpl_binds _ env [] = return env
230 simpl_binds dump env (bind:binds) = do { env' <- trace_bind dump bind $
232 ; simpl_binds dump env' binds }
234 trace_bind True bind = pprTrace "SimplBind" (ppr (bindersOf bind))
235 trace_bind False _ = \x -> x
237 simpl_bind env (Rec pairs) = simplRecBind env TopLevel pairs
238 simpl_bind env (NonRec b r) = simplRecOrTopPair env' TopLevel b b' r
240 (env', b') = addBndrRules env b (lookupRecBndr env b)
244 %************************************************************************
246 \subsection{Lazy bindings}
248 %************************************************************************
250 simplRecBind is used for
251 * recursive bindings only
254 simplRecBind :: SimplEnv -> TopLevelFlag
257 simplRecBind env0 top_lvl pairs0
258 = do { let (env_with_info, triples) = mapAccumL add_rules env0 pairs0
259 ; env1 <- go (zapFloats env_with_info) triples
260 ; return (env0 `addRecFloats` env1) }
261 -- addFloats adds the floats from env1,
262 -- _and_ updates env0 with the in-scope set from env1
264 add_rules :: SimplEnv -> (InBndr,InExpr) -> (SimplEnv, (InBndr, OutBndr, InExpr))
265 -- Add the (substituted) rules to the binder
266 add_rules env (bndr, rhs) = (env', (bndr, bndr', rhs))
268 (env', bndr') = addBndrRules env bndr (lookupRecBndr env bndr)
270 go env [] = return env
272 go env ((old_bndr, new_bndr, rhs) : pairs)
273 = do { env' <- simplRecOrTopPair env top_lvl old_bndr new_bndr rhs
277 simplOrTopPair is used for
278 * recursive bindings (whether top level or not)
279 * top-level non-recursive bindings
281 It assumes the binder has already been simplified, but not its IdInfo.
284 simplRecOrTopPair :: SimplEnv
286 -> InId -> OutBndr -> InExpr -- Binder and rhs
287 -> SimplM SimplEnv -- Returns an env that includes the binding
289 simplRecOrTopPair env top_lvl old_bndr new_bndr rhs
290 | preInlineUnconditionally env top_lvl old_bndr rhs -- Check for unconditional inline
291 = do { tick (PreInlineUnconditionally old_bndr)
292 ; return (extendIdSubst env old_bndr (mkContEx env rhs)) }
295 = simplLazyBind env top_lvl Recursive old_bndr new_bndr rhs env
296 -- May not actually be recursive, but it doesn't matter
300 simplLazyBind is used for
301 * [simplRecOrTopPair] recursive bindings (whether top level or not)
302 * [simplRecOrTopPair] top-level non-recursive bindings
303 * [simplNonRecE] non-top-level *lazy* non-recursive bindings
306 1. It assumes that the binder is *already* simplified,
307 and is in scope, and its IdInfo too, except unfolding
309 2. It assumes that the binder type is lifted.
311 3. It does not check for pre-inline-unconditionallly;
312 that should have been done already.
315 simplLazyBind :: SimplEnv
316 -> TopLevelFlag -> RecFlag
317 -> InId -> OutId -- Binder, both pre-and post simpl
318 -- The OutId has IdInfo, except arity, unfolding
319 -> InExpr -> SimplEnv -- The RHS and its environment
322 simplLazyBind env top_lvl is_rec bndr bndr1 rhs rhs_se
323 = do { let rhs_env = rhs_se `setInScope` env
324 (tvs, body) = case collectTyBinders rhs of
325 (tvs, body) | not_lam body -> (tvs,body)
326 | otherwise -> ([], rhs)
327 not_lam (Lam _ _) = False
329 -- Do not do the "abstract tyyvar" thing if there's
330 -- a lambda inside, becuase it defeats eta-reduction
331 -- f = /\a. \x. g a x
334 ; (body_env, tvs') <- simplBinders rhs_env tvs
335 -- See Note [Floating and type abstraction] in SimplUtils
338 ; (body_env1, body1) <- simplExprF body_env body mkRhsStop
339 -- ANF-ise a constructor or PAP rhs
340 ; (body_env2, body2) <- prepareRhs body_env1 bndr1 body1
343 <- if not (doFloatFromRhs top_lvl is_rec False body2 body_env2)
344 then -- No floating, just wrap up!
345 do { rhs' <- mkLam env tvs' (wrapFloats body_env2 body2)
346 ; return (env, rhs') }
348 else if null tvs then -- Simple floating
349 do { tick LetFloatFromLet
350 ; return (addFloats env body_env2, body2) }
352 else -- Do type-abstraction first
353 do { tick LetFloatFromLet
354 ; (poly_binds, body3) <- abstractFloats tvs' body_env2 body2
355 ; rhs' <- mkLam env tvs' body3
356 ; env' <- foldlM (addPolyBind top_lvl) env poly_binds
357 ; return (env', rhs') }
359 ; completeBind env' top_lvl bndr bndr1 rhs' }
362 A specialised variant of simplNonRec used when the RHS is already simplified,
363 notably in knownCon. It uses case-binding where necessary.
366 simplNonRecX :: SimplEnv
367 -> InId -- Old binder
368 -> OutExpr -- Simplified RHS
371 simplNonRecX env bndr new_rhs
372 | isDeadBinder bndr -- Not uncommon; e.g. case (a,b) of b { (p,q) -> p }
373 = return env -- Here b is dead, and we avoid creating
374 | otherwise -- the binding b = (a,b)
375 = do { (env', bndr') <- simplBinder env bndr
376 ; completeNonRecX env' (isStrictId bndr) bndr bndr' new_rhs }
378 completeNonRecX :: SimplEnv
380 -> InId -- Old binder
381 -> OutId -- New binder
382 -> OutExpr -- Simplified RHS
385 completeNonRecX env is_strict old_bndr new_bndr new_rhs
386 = do { (env1, rhs1) <- prepareRhs (zapFloats env) new_bndr new_rhs
388 if doFloatFromRhs NotTopLevel NonRecursive is_strict rhs1 env1
389 then do { tick LetFloatFromLet
390 ; return (addFloats env env1, rhs1) } -- Add the floats to the main env
391 else return (env, wrapFloats env1 rhs1) -- Wrap the floats around the RHS
392 ; completeBind env2 NotTopLevel old_bndr new_bndr rhs2 }
395 {- No, no, no! Do not try preInlineUnconditionally in completeNonRecX
396 Doing so risks exponential behaviour, because new_rhs has been simplified once already
397 In the cases described by the folowing commment, postInlineUnconditionally will
398 catch many of the relevant cases.
399 -- This happens; for example, the case_bndr during case of
400 -- known constructor: case (a,b) of x { (p,q) -> ... }
401 -- Here x isn't mentioned in the RHS, so we don't want to
402 -- create the (dead) let-binding let x = (a,b) in ...
404 -- Similarly, single occurrences can be inlined vigourously
405 -- e.g. case (f x, g y) of (a,b) -> ....
406 -- If a,b occur once we can avoid constructing the let binding for them.
408 Furthermore in the case-binding case preInlineUnconditionally risks extra thunks
409 -- Consider case I# (quotInt# x y) of
410 -- I# v -> let w = J# v in ...
411 -- If we gaily inline (quotInt# x y) for v, we end up building an
413 -- let w = J# (quotInt# x y) in ...
414 -- because quotInt# can fail.
416 | preInlineUnconditionally env NotTopLevel bndr new_rhs
417 = thing_inside (extendIdSubst env bndr (DoneEx new_rhs))
420 ----------------------------------
421 prepareRhs takes a putative RHS, checks whether it's a PAP or
422 constructor application and, if so, converts it to ANF, so that the
423 resulting thing can be inlined more easily. Thus
430 We also want to deal well cases like this
431 v = (f e1 `cast` co) e2
432 Here we want to make e1,e2 trivial and get
433 x1 = e1; x2 = e2; v = (f x1 `cast` co) v2
434 That's what the 'go' loop in prepareRhs does
437 prepareRhs :: SimplEnv -> OutId -> OutExpr -> SimplM (SimplEnv, OutExpr)
438 -- Adds new floats to the env iff that allows us to return a good RHS
439 prepareRhs env id (Cast rhs co) -- Note [Float coercions]
440 | (ty1, _ty2) <- coercionKind co -- Do *not* do this if rhs has an unlifted type
441 , not (isUnLiftedType ty1) -- see Note [Float coercions (unlifted)]
442 = do { (env', rhs') <- makeTrivialWithInfo env sanitised_info rhs
443 ; return (env', Cast rhs' co) }
445 sanitised_info = vanillaIdInfo `setStrictnessInfo` strictnessInfo info
446 `setDemandInfo` demandInfo info
449 prepareRhs env0 _ rhs0
450 = do { (_is_exp, env1, rhs1) <- go 0 env0 rhs0
451 ; return (env1, rhs1) }
453 go n_val_args env (Cast rhs co)
454 = do { (is_exp, env', rhs') <- go n_val_args env rhs
455 ; return (is_exp, env', Cast rhs' co) }
456 go n_val_args env (App fun (Type ty))
457 = do { (is_exp, env', rhs') <- go n_val_args env fun
458 ; return (is_exp, env', App rhs' (Type ty)) }
459 go n_val_args env (App fun arg)
460 = do { (is_exp, env', fun') <- go (n_val_args+1) env fun
462 True -> do { (env'', arg') <- makeTrivial env' arg
463 ; return (True, env'', App fun' arg') }
464 False -> return (False, env, App fun arg) }
465 go n_val_args env (Var fun)
466 = return (is_exp, env, Var fun)
468 is_exp = isExpandableApp fun n_val_args -- The fun a constructor or PAP
469 -- See Note [CONLIKE pragma] in BasicTypes
470 -- The definition of is_exp should match that in
471 -- OccurAnal.occAnalApp
474 = return (False, env, other)
478 Note [Float coercions]
479 ~~~~~~~~~~~~~~~~~~~~~~
480 When we find the binding
482 we'd like to transform it to
484 x = x `cast` co -- A trivial binding
485 There's a chance that e will be a constructor application or function, or something
486 like that, so moving the coerion to the usage site may well cancel the coersions
487 and lead to further optimisation. Example:
490 data instance T Int = T Int
492 foo :: Int -> Int -> Int
497 go n = case x of { T m -> go (n-m) }
498 -- This case should optimise
500 Note [Preserve strictness when floating coercions]
501 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
502 In the Note [Float coercions] transformation, keep the strictness info.
504 f = e `cast` co -- f has strictness SSL
506 f' = e -- f' also has strictness SSL
507 f = f' `cast` co -- f still has strictness SSL
509 Its not wrong to drop it on the floor, but better to keep it.
511 Note [Float coercions (unlifted)]
512 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
513 BUT don't do [Float coercions] if 'e' has an unlifted type.
516 foo :: Int = (error (# Int,Int #) "urk")
517 `cast` CoUnsafe (# Int,Int #) Int
519 If do the makeTrivial thing to the error call, we'll get
520 foo = case error (# Int,Int #) "urk" of v -> v `cast` ...
521 But 'v' isn't in scope!
523 These strange casts can happen as a result of case-of-case
524 bar = case (case x of { T -> (# 2,3 #); F -> error "urk" }) of
529 makeTrivial :: SimplEnv -> OutExpr -> SimplM (SimplEnv, OutExpr)
530 -- Binds the expression to a variable, if it's not trivial, returning the variable
531 makeTrivial env expr = makeTrivialWithInfo env vanillaIdInfo expr
533 makeTrivialWithInfo :: SimplEnv -> IdInfo -> OutExpr -> SimplM (SimplEnv, OutExpr)
534 -- Propagate strictness and demand info to the new binder
535 -- Note [Preserve strictness when floating coercions]
536 makeTrivialWithInfo env info expr
539 | otherwise -- See Note [Take care] below
540 = do { uniq <- getUniqueM
541 ; let name = mkSystemVarName uniq (fsLit "a")
542 var = mkLocalIdWithInfo name (exprType expr) info
543 ; env' <- completeNonRecX env False var var expr
544 ; return (env', substExpr env' (Var var)) }
545 -- The substitution is needed becase we're constructing a new binding
547 -- And if rhs is of form (rhs1 |> co), then we might get
550 -- and now a's RHS is trivial and can be substituted out, and that
551 -- is what completeNonRecX will do
555 %************************************************************************
557 \subsection{Completing a lazy binding}
559 %************************************************************************
562 * deals only with Ids, not TyVars
563 * takes an already-simplified binder and RHS
564 * is used for both recursive and non-recursive bindings
565 * is used for both top-level and non-top-level bindings
567 It does the following:
568 - tries discarding a dead binding
569 - tries PostInlineUnconditionally
570 - add unfolding [this is the only place we add an unfolding]
573 It does *not* attempt to do let-to-case. Why? Because it is used for
574 - top-level bindings (when let-to-case is impossible)
575 - many situations where the "rhs" is known to be a WHNF
576 (so let-to-case is inappropriate).
578 Nor does it do the atomic-argument thing
581 completeBind :: SimplEnv
582 -> TopLevelFlag -- Flag stuck into unfolding
583 -> InId -- Old binder
584 -> OutId -> OutExpr -- New binder and RHS
586 -- completeBind may choose to do its work
587 -- * by extending the substitution (e.g. let x = y in ...)
588 -- * or by adding to the floats in the envt
590 completeBind env top_lvl old_bndr new_bndr new_rhs
591 = do { let old_info = idInfo old_bndr
592 old_unf = unfoldingInfo old_info
593 occ_info = occInfo old_info
595 ; new_unfolding <- simplUnfolding env top_lvl old_bndr occ_info new_rhs old_unf
597 ; if postInlineUnconditionally env top_lvl new_bndr occ_info new_rhs new_unfolding
598 -- Inline and discard the binding
599 then do { tick (PostInlineUnconditionally old_bndr)
600 ; -- pprTrace "postInlineUnconditionally" (ppr old_bndr <+> equals <+> ppr new_rhs) $
601 return (extendIdSubst env old_bndr (DoneEx new_rhs)) }
602 -- Use the substitution to make quite, quite sure that the
603 -- substitution will happen, since we are going to discard the binding
605 else return (addNonRecWithUnf env new_bndr new_rhs new_unfolding) }
607 ------------------------------
608 addPolyBind :: TopLevelFlag -> SimplEnv -> OutBind -> SimplM SimplEnv
609 -- Add a new binding to the environment, complete with its unfolding
610 -- but *do not* do postInlineUnconditionally, because we have already
611 -- processed some of the scope of the binding
612 -- We still want the unfolding though. Consider
614 -- x = /\a. let y = ... in Just y
616 -- Then we float the y-binding out (via abstractFloats and addPolyBind)
617 -- but 'x' may well then be inlined in 'body' in which case we'd like the
618 -- opportunity to inline 'y' too.
620 addPolyBind top_lvl env (NonRec poly_id rhs)
621 = do { unfolding <- simplUnfolding env top_lvl poly_id NoOccInfo rhs noUnfolding
622 -- Assumes that poly_id did not have an INLINE prag
623 -- which is perhaps wrong. ToDo: think about this
624 ; return (addNonRecWithUnf env poly_id rhs unfolding) }
626 addPolyBind _ env bind@(Rec _) = return (extendFloats env bind)
627 -- Hack: letrecs are more awkward, so we extend "by steam"
628 -- without adding unfoldings etc. At worst this leads to
629 -- more simplifier iterations
631 ------------------------------
632 addNonRecWithUnf :: SimplEnv
633 -> OutId -> OutExpr -- New binder and RHS
634 -> Unfolding -- New unfolding
636 addNonRecWithUnf env new_bndr new_rhs new_unfolding
637 = let new_arity = exprArity new_rhs
638 old_arity = idArity new_bndr
639 info1 = idInfo new_bndr `setArityInfo` new_arity
641 -- Unfolding info: Note [Setting the new unfolding]
642 info2 = info1 `setUnfoldingInfo` new_unfolding
644 -- Demand info: Note [Setting the demand info]
645 info3 | isEvaldUnfolding new_unfolding = zapDemandInfo info2 `orElse` info2
648 final_id = new_bndr `setIdInfo` info3
649 dmd_arity = length $ fst $ splitStrictSig $ idStrictness new_bndr
651 ASSERT( isId new_bndr )
652 WARN( new_arity < old_arity || new_arity < dmd_arity,
653 (ptext (sLit "Arity decrease:") <+> ppr final_id <+> ppr old_arity
654 <+> ppr new_arity <+> ppr dmd_arity) )
655 -- Note [Arity decrease]
657 final_id `seq` -- This seq forces the Id, and hence its IdInfo,
658 -- and hence any inner substitutions
659 -- pprTrace "Binding" (ppr final_id <+> ppr unfolding) $
660 addNonRec env final_id new_rhs
661 -- The addNonRec adds it to the in-scope set too
663 ------------------------------
664 simplUnfolding :: SimplEnv-> TopLevelFlag
666 -> OccInfo -> OutExpr
667 -> Unfolding -> SimplM Unfolding
668 -- Note [Setting the new unfolding]
669 simplUnfolding env _ _ _ _ (DFunUnfolding con ops)
670 = return (DFunUnfolding con ops')
672 ops' = map (CoreSubst.substExpr (mkCoreSubst env)) ops
674 simplUnfolding env top_lvl id _ _
675 (CoreUnfolding { uf_tmpl = expr, uf_arity = arity
676 , uf_src = src, uf_guidance = guide })
677 | isInlineRuleSource src
678 = do { expr' <- simplExpr rule_env expr
679 ; let src' = CoreSubst.substUnfoldingSource (mkCoreSubst env) src
680 ; return (mkCoreUnfolding (isTopLevel top_lvl) src' expr' arity guide) }
681 -- See Note [Top-level flag on inline rules] in CoreUnfold
683 rule_env = updMode (updModeForInlineRules (idInlineActivation id)) env
684 -- See Note [Simplifying gently inside InlineRules] in SimplUtils
686 simplUnfolding _ top_lvl id _occ_info new_rhs _
687 = return (mkUnfolding (isTopLevel top_lvl) (isBottomingId id) new_rhs)
688 -- We make an unfolding *even for loop-breakers*.
689 -- Reason: (a) It might be useful to know that they are WHNF
690 -- (b) In TidyPgm we currently assume that, if we want to
691 -- expose the unfolding then indeed we *have* an unfolding
692 -- to expose. (We could instead use the RHS, but currently
693 -- we don't.) The simple thing is always to have one.
696 Note [Arity decrease]
697 ~~~~~~~~~~~~~~~~~~~~~
698 Generally speaking the arity of a binding should not decrease. But it *can*
699 legitimately happen becuase of RULES. Eg
701 where g has arity 2, will have arity 2. But if there's a rewrite rule
703 where h has arity 1, then f's arity will decrease. Here's a real-life example,
704 which is in the output of Specialise:
707 $dm {Arity 2} = \d.\x. op d
708 {-# RULES forall d. $dm Int d = $s$dm #-}
710 dInt = MkD .... opInt ...
711 opInt {Arity 1} = $dm dInt
713 $s$dm {Arity 0} = \x. op dInt }
715 Here opInt has arity 1; but when we apply the rule its arity drops to 0.
716 That's why Specialise goes to a little trouble to pin the right arity
717 on specialised functions too.
719 Note [Setting the new unfolding]
720 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
721 * If there's an INLINE pragma, we simplify the RHS gently. Maybe we
722 should do nothing at all, but simplifying gently might get rid of
725 * If not, we make an unfolding from the new RHS. But *only* for
726 non-loop-breakers. Making loop breakers not have an unfolding at all
727 means that we can avoid tests in exprIsConApp, for example. This is
728 important: if exprIsConApp says 'yes' for a recursive thing, then we
729 can get into an infinite loop
731 If there's an InlineRule on a loop breaker, we hang on to the inlining.
732 It's pretty dodgy, but the user did say 'INLINE'. May need to revisit
735 Note [Setting the demand info]
736 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
737 If the unfolding is a value, the demand info may
738 go pear-shaped, so we nuke it. Example:
740 case x of (p,q) -> h p q x
741 Here x is certainly demanded. But after we've nuked
742 the case, we'll get just
743 let x = (a,b) in h a b x
744 and now x is not demanded (I'm assuming h is lazy)
745 This really happens. Similarly
746 let f = \x -> e in ...f..f...
747 After inlining f at some of its call sites the original binding may
748 (for example) be no longer strictly demanded.
749 The solution here is a bit ad hoc...
752 %************************************************************************
754 \subsection[Simplify-simplExpr]{The main function: simplExpr}
756 %************************************************************************
758 The reason for this OutExprStuff stuff is that we want to float *after*
759 simplifying a RHS, not before. If we do so naively we get quadratic
760 behaviour as things float out.
762 To see why it's important to do it after, consider this (real) example:
776 a -- Can't inline a this round, cos it appears twice
780 Each of the ==> steps is a round of simplification. We'd save a
781 whole round if we float first. This can cascade. Consider
786 let f = let d1 = ..d.. in \y -> e
790 in \x -> ...(\y ->e)...
792 Only in this second round can the \y be applied, and it
793 might do the same again.
797 simplExpr :: SimplEnv -> CoreExpr -> SimplM CoreExpr
798 simplExpr env expr = simplExprC env expr mkBoringStop
800 simplExprC :: SimplEnv -> CoreExpr -> SimplCont -> SimplM CoreExpr
801 -- Simplify an expression, given a continuation
802 simplExprC env expr cont
803 = -- pprTrace "simplExprC" (ppr expr $$ ppr cont {- $$ ppr (seIdSubst env) -} $$ ppr (seFloats env) ) $
804 do { (env', expr') <- simplExprF (zapFloats env) expr cont
805 ; -- pprTrace "simplExprC ret" (ppr expr $$ ppr expr') $
806 -- pprTrace "simplExprC ret3" (ppr (seInScope env')) $
807 -- pprTrace "simplExprC ret4" (ppr (seFloats env')) $
808 return (wrapFloats env' expr') }
810 --------------------------------------------------
811 simplExprF :: SimplEnv -> InExpr -> SimplCont
812 -> SimplM (SimplEnv, OutExpr)
814 simplExprF env e cont
815 = -- pprTrace "simplExprF" (ppr e $$ ppr cont $$ ppr (seTvSubst env) $$ ppr (seIdSubst env) {- $$ ppr (seFloats env) -} ) $
816 simplExprF' env e cont
818 simplExprF' :: SimplEnv -> InExpr -> SimplCont
819 -> SimplM (SimplEnv, OutExpr)
820 simplExprF' env (Var v) cont = simplVar env v cont
821 simplExprF' env (Lit lit) cont = rebuild env (Lit lit) cont
822 simplExprF' env (Note n expr) cont = simplNote env n expr cont
823 simplExprF' env (Cast body co) cont = simplCast env body co cont
824 simplExprF' env (App fun arg) cont = simplExprF env fun $
825 ApplyTo NoDup arg env cont
827 simplExprF' env expr@(Lam _ _) cont
828 = simplLam env (map zap bndrs) body cont
829 -- The main issue here is under-saturated lambdas
830 -- (\x1. \x2. e) arg1
831 -- Here x1 might have "occurs-once" occ-info, because occ-info
832 -- is computed assuming that a group of lambdas is applied
833 -- all at once. If there are too few args, we must zap the
836 n_args = countArgs cont
837 n_params = length bndrs
838 (bndrs, body) = collectBinders expr
839 zap | n_args >= n_params = \b -> b
840 | otherwise = \b -> if isTyVar b then b
842 -- NB: we count all the args incl type args
843 -- so we must count all the binders (incl type lambdas)
845 simplExprF' env (Type ty) cont
846 = ASSERT( contIsRhsOrArg cont )
847 do { ty' <- simplCoercion env ty
848 ; rebuild env (Type ty') cont }
850 simplExprF' env (Case scrut bndr _ alts) cont
851 | not (switchIsOn (getSwitchChecker env) NoCaseOfCase)
852 = -- Simplify the scrutinee with a Select continuation
853 simplExprF env scrut (Select NoDup bndr alts env cont)
856 = -- If case-of-case is off, simply simplify the case expression
857 -- in a vanilla Stop context, and rebuild the result around it
858 do { case_expr' <- simplExprC env scrut case_cont
859 ; rebuild env case_expr' cont }
861 case_cont = Select NoDup bndr alts env mkBoringStop
863 simplExprF' env (Let (Rec pairs) body) cont
864 = do { env' <- simplRecBndrs env (map fst pairs)
865 -- NB: bndrs' don't have unfoldings or rules
866 -- We add them as we go down
868 ; env'' <- simplRecBind env' NotTopLevel pairs
869 ; simplExprF env'' body cont }
871 simplExprF' env (Let (NonRec bndr rhs) body) cont
872 = simplNonRecE env bndr (rhs, env) ([], body) cont
874 ---------------------------------
875 simplType :: SimplEnv -> InType -> SimplM OutType
876 -- Kept monadic just so we can do the seqType
878 = -- pprTrace "simplType" (ppr ty $$ ppr (seTvSubst env)) $
879 seqType new_ty `seq` return new_ty
881 new_ty = substTy env ty
883 ---------------------------------
884 simplCoercion :: SimplEnv -> InType -> SimplM OutType
885 -- The InType isn't *necessarily* a coercion, but it might be
886 -- (in a type application, say) and optCoercion is a no-op on types
888 = seqType new_co `seq` return new_co
890 new_co = optCoercion (getTvSubst env) co
894 %************************************************************************
896 \subsection{The main rebuilder}
898 %************************************************************************
901 rebuild :: SimplEnv -> OutExpr -> SimplCont -> SimplM (SimplEnv, OutExpr)
902 -- At this point the substitution in the SimplEnv should be irrelevant
903 -- only the in-scope set and floats should matter
904 rebuild env expr cont0
905 = -- pprTrace "rebuild" (ppr expr $$ ppr cont0 $$ ppr (seFloats env)) $
907 Stop {} -> return (env, expr)
908 CoerceIt co cont -> rebuild env (mkCoerce co expr) cont
909 Select _ bndr alts se cont -> rebuildCase (se `setFloats` env) expr bndr alts cont
910 StrictArg info _ cont -> rebuildCall env (info `addArgTo` expr) cont
911 StrictBind b bs body se cont -> do { env' <- simplNonRecX (se `setFloats` env) b expr
912 ; simplLam env' bs body cont }
913 ApplyTo _ arg se cont -> do { arg' <- simplExpr (se `setInScope` env) arg
914 ; rebuild env (App expr arg') cont }
918 %************************************************************************
922 %************************************************************************
925 simplCast :: SimplEnv -> InExpr -> Coercion -> SimplCont
926 -> SimplM (SimplEnv, OutExpr)
927 simplCast env body co0 cont0
928 = do { co1 <- simplCoercion env co0
929 ; simplExprF env body (addCoerce co1 cont0) }
931 addCoerce co cont = add_coerce co (coercionKind co) cont
933 add_coerce _co (s1, k1) cont -- co :: ty~ty
934 | s1 `coreEqType` k1 = cont -- is a no-op
936 add_coerce co1 (s1, _k2) (CoerceIt co2 cont)
937 | (_l1, t1) <- coercionKind co2
938 -- e |> (g1 :: S1~L) |> (g2 :: L~T1)
941 -- e |> (g1 . g2 :: S1~T1) otherwise
943 -- For example, in the initial form of a worker
944 -- we may find (coerce T (coerce S (\x.e))) y
945 -- and we'd like it to simplify to e[y/x] in one round
947 , s1 `coreEqType` t1 = cont -- The coerces cancel out
948 | otherwise = CoerceIt (mkTransCoercion co1 co2) cont
950 add_coerce co (s1s2, _t1t2) (ApplyTo dup (Type arg_ty) arg_se cont)
951 -- (f |> g) ty ---> (f ty) |> (g @ ty)
952 -- This implements the PushT and PushC rules from the paper
953 | Just (tyvar,_) <- splitForAllTy_maybe s1s2
955 (new_arg_ty, new_cast)
956 | isCoVar tyvar = (new_arg_co, mkCselRCoercion co) -- PushC rule
957 | otherwise = (ty', mkInstCoercion co ty') -- PushT rule
959 ApplyTo dup (Type new_arg_ty) (zapSubstEnv arg_se) (addCoerce new_cast cont)
961 ty' = substTy (arg_se `setInScope` env) arg_ty
962 new_arg_co = mkCsel1Coercion co `mkTransCoercion`
963 ty' `mkTransCoercion`
964 mkSymCoercion (mkCsel2Coercion co)
966 add_coerce co (s1s2, _t1t2) (ApplyTo dup arg arg_se cont)
967 | not (isTypeArg arg) -- This implements the Push rule from the paper
968 , isFunTy s1s2 -- t1t2 must be a function type, becuase it's applied
969 -- (e |> (g :: s1s2 ~ t1->t2)) f
971 -- (e (f |> (arg g :: t1~s1))
972 -- |> (res g :: s2->t2)
974 -- t1t2 must be a function type, t1->t2, because it's applied
975 -- to something but s1s2 might conceivably not be
977 -- When we build the ApplyTo we can't mix the out-types
978 -- with the InExpr in the argument, so we simply substitute
979 -- to make it all consistent. It's a bit messy.
980 -- But it isn't a common case.
982 -- Example of use: Trac #995
983 = ApplyTo dup new_arg (zapSubstEnv arg_se) (addCoerce co2 cont)
985 -- we split coercion t1->t2 ~ s1->s2 into t1 ~ s1 and
986 -- t2 ~ s2 with left and right on the curried form:
987 -- (->) t1 t2 ~ (->) s1 s2
988 [co1, co2] = decomposeCo 2 co
989 new_arg = mkCoerce (mkSymCoercion co1) arg'
990 arg' = substExpr (arg_se `setInScope` env) arg
992 add_coerce co _ cont = CoerceIt co cont
996 %************************************************************************
1000 %************************************************************************
1003 simplLam :: SimplEnv -> [InId] -> InExpr -> SimplCont
1004 -> SimplM (SimplEnv, OutExpr)
1006 simplLam env [] body cont = simplExprF env body cont
1009 simplLam env (bndr:bndrs) body (ApplyTo _ arg arg_se cont)
1010 = do { tick (BetaReduction bndr)
1011 ; simplNonRecE env bndr (arg, arg_se) (bndrs, body) cont }
1013 -- Not enough args, so there are real lambdas left to put in the result
1014 simplLam env bndrs body cont
1015 = do { (env', bndrs') <- simplLamBndrs env bndrs
1016 ; body' <- simplExpr env' body
1017 ; new_lam <- mkLam env' bndrs' body'
1018 ; rebuild env' new_lam cont }
1021 simplNonRecE :: SimplEnv
1022 -> InBndr -- The binder
1023 -> (InExpr, SimplEnv) -- Rhs of binding (or arg of lambda)
1024 -> ([InBndr], InExpr) -- Body of the let/lambda
1027 -> SimplM (SimplEnv, OutExpr)
1029 -- simplNonRecE is used for
1030 -- * non-top-level non-recursive lets in expressions
1033 -- It deals with strict bindings, via the StrictBind continuation,
1034 -- which may abort the whole process
1036 -- The "body" of the binding comes as a pair of ([InId],InExpr)
1037 -- representing a lambda; so we recurse back to simplLam
1038 -- Why? Because of the binder-occ-info-zapping done before
1039 -- the call to simplLam in simplExprF (Lam ...)
1041 -- First deal with type applications and type lets
1042 -- (/\a. e) (Type ty) and (let a = Type ty in e)
1043 simplNonRecE env bndr (Type ty_arg, rhs_se) (bndrs, body) cont
1044 = ASSERT( isTyVar bndr )
1045 do { ty_arg' <- simplType (rhs_se `setInScope` env) ty_arg
1046 ; simplLam (extendTvSubst env bndr ty_arg') bndrs body cont }
1048 simplNonRecE env bndr (rhs, rhs_se) (bndrs, body) cont
1049 | preInlineUnconditionally env NotTopLevel bndr rhs
1050 = do { tick (PreInlineUnconditionally bndr)
1051 ; simplLam (extendIdSubst env bndr (mkContEx rhs_se rhs)) bndrs body cont }
1054 = do { simplExprF (rhs_se `setFloats` env) rhs
1055 (StrictBind bndr bndrs body env cont) }
1058 = ASSERT( not (isTyVar bndr) )
1059 do { (env1, bndr1) <- simplNonRecBndr env bndr
1060 ; let (env2, bndr2) = addBndrRules env1 bndr bndr1
1061 ; env3 <- simplLazyBind env2 NotTopLevel NonRecursive bndr bndr2 rhs rhs_se
1062 ; simplLam env3 bndrs body cont }
1066 %************************************************************************
1070 %************************************************************************
1073 -- Hack alert: we only distinguish subsumed cost centre stacks for the
1074 -- purposes of inlining. All other CCCSs are mapped to currentCCS.
1075 simplNote :: SimplEnv -> Note -> CoreExpr -> SimplCont
1076 -> SimplM (SimplEnv, OutExpr)
1077 simplNote env (SCC cc) e cont
1078 | pushCCisNop cc (getEnclosingCC env) -- scc "f" (...(scc "f" e)...)
1079 = simplExprF env e cont -- ==> scc "f" (...e...)
1081 = do { e' <- simplExpr (setEnclosingCC env currentCCS) e
1082 ; rebuild env (mkSCC cc e') cont }
1084 simplNote env (CoreNote s) e cont
1085 = do { e' <- simplExpr env e
1086 ; rebuild env (Note (CoreNote s) e') cont }
1090 %************************************************************************
1092 \subsection{Dealing with calls}
1094 %************************************************************************
1097 simplVar :: SimplEnv -> Id -> SimplCont -> SimplM (SimplEnv, OutExpr)
1098 simplVar env var cont
1099 = case substId env var of
1100 DoneEx e -> simplExprF (zapSubstEnv env) e cont
1101 ContEx tvs ids e -> simplExprF (setSubstEnv env tvs ids) e cont
1102 DoneId var1 -> completeCall env var1 cont
1103 -- Note [zapSubstEnv]
1104 -- The template is already simplified, so don't re-substitute.
1105 -- This is VITAL. Consider
1107 -- let y = \z -> ...x... in
1109 -- We'll clone the inner \x, adding x->x' in the id_subst
1110 -- Then when we inline y, we must *not* replace x by x' in
1111 -- the inlined copy!!
1113 ---------------------------------------------------------
1114 -- Dealing with a call site
1116 completeCall :: SimplEnv -> Id -> SimplCont -> SimplM (SimplEnv, OutExpr)
1117 completeCall env var cont
1118 = do { ------------- Try inlining ----------------
1119 dflags <- getDOptsSmpl
1120 ; let (args,call_cont) = contArgs cont
1121 -- The args are OutExprs, obtained by *lazily* substituting
1122 -- in the args found in cont. These args are only examined
1123 -- to limited depth (unless a rule fires). But we must do
1124 -- the substitution; rule matching on un-simplified args would
1127 arg_infos = [interestingArg arg | arg <- args, isValArg arg]
1128 n_val_args = length arg_infos
1129 interesting_cont = interestingCallContext call_cont
1130 unfolding = activeUnfolding env var
1131 maybe_inline = callSiteInline dflags var unfolding
1132 (null args) arg_infos interesting_cont
1133 ; case maybe_inline of {
1134 Just unfolding -- There is an inlining!
1135 -> do { tick (UnfoldingDone var)
1136 ; (if dopt Opt_D_dump_inlinings dflags then
1137 pprTrace ("Inlining done: " ++ showSDoc (ppr var)) (vcat [
1138 text "Before:" <+> ppr var <+> sep (map pprParendExpr args),
1139 text "Inlined fn: " <+> nest 2 (ppr unfolding),
1140 text "Cont: " <+> ppr call_cont])
1143 simplExprF (zapSubstEnv env) unfolding cont }
1145 ; Nothing -> do -- No inlining!
1147 { rule_base <- getSimplRules
1148 ; let info = mkArgInfo var (getRules rule_base var) n_val_args call_cont
1149 ; rebuildCall env info cont
1152 rebuildCall :: SimplEnv
1155 -> SimplM (SimplEnv, OutExpr)
1156 rebuildCall env (ArgInfo { ai_fun = fun, ai_args = rev_args, ai_strs = [] }) cont
1157 -- When we run out of strictness args, it means
1158 -- that the call is definitely bottom; see SimplUtils.mkArgInfo
1159 -- Then we want to discard the entire strict continuation. E.g.
1160 -- * case (error "hello") of { ... }
1161 -- * (error "Hello") arg
1162 -- * f (error "Hello") where f is strict
1164 -- Then, especially in the first of these cases, we'd like to discard
1165 -- the continuation, leaving just the bottoming expression. But the
1166 -- type might not be right, so we may have to add a coerce.
1167 | not (contIsTrivial cont) -- Only do this if there is a non-trivial
1168 = return (env, mk_coerce res) -- contination to discard, else we do it
1169 where -- again and again!
1170 res = mkApps (Var fun) (reverse rev_args)
1171 res_ty = exprType res
1172 cont_ty = contResultType env res_ty cont
1173 co = mkUnsafeCoercion res_ty cont_ty
1174 mk_coerce expr | cont_ty `coreEqType` res_ty = expr
1175 | otherwise = mkCoerce co expr
1177 rebuildCall env info (ApplyTo _ (Type arg_ty) se cont)
1178 = do { ty' <- simplCoercion (se `setInScope` env) arg_ty
1179 ; rebuildCall env (info `addArgTo` Type ty') cont }
1181 rebuildCall env info@(ArgInfo { ai_encl = encl_rules
1182 , ai_strs = str:strs, ai_discs = disc:discs })
1183 (ApplyTo _ arg arg_se cont)
1184 | str -- Strict argument
1185 = -- pprTrace "Strict Arg" (ppr arg $$ ppr (seIdSubst env) $$ ppr (seInScope env)) $
1186 simplExprF (arg_se `setFloats` env) arg
1187 (StrictArg info' cci cont)
1190 | otherwise -- Lazy argument
1191 -- DO NOT float anything outside, hence simplExprC
1192 -- There is no benefit (unlike in a let-binding), and we'd
1193 -- have to be very careful about bogus strictness through
1194 -- floating a demanded let.
1195 = do { arg' <- simplExprC (arg_se `setInScope` env) arg
1197 ; rebuildCall env (addArgTo info' arg') cont }
1199 info' = info { ai_strs = strs, ai_discs = discs }
1200 cci | encl_rules || disc > 0 = ArgCtxt encl_rules -- Be keener here
1201 | otherwise = BoringCtxt -- Nothing interesting
1203 rebuildCall env (ArgInfo { ai_fun = fun, ai_args = rev_args, ai_rules = rules }) cont
1204 = do { -- We've accumulated a simplified call in <fun,rev_args>
1205 -- so try rewrite rules; see Note [RULEs apply to simplified arguments]
1206 -- See also Note [Rules for recursive functions]
1207 ; let args = reverse rev_args
1208 env' = zapSubstEnv env
1209 ; mb_rule <- tryRules env rules fun args cont
1211 Just (n_args, rule_rhs) -> simplExprF env' rule_rhs $
1212 pushArgs env' (drop n_args args) cont ;
1213 -- n_args says how many args the rule consumed
1214 ; Nothing -> rebuild env (mkApps (Var fun) args) cont -- No rules
1218 Note [RULES apply to simplified arguments]
1219 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1220 It's very desirable to try RULES once the arguments have been simplified, because
1221 doing so ensures that rule cascades work in one pass. Consider
1222 {-# RULES g (h x) = k x
1225 Then we want to rewrite (g (h x)) to (k x) and only then try f's rules. If
1226 we match f's rules against the un-simplified RHS, it won't match. This
1227 makes a particularly big difference when superclass selectors are involved:
1228 op ($p1 ($p2 (df d)))
1229 We want all this to unravel in one sweeep.
1233 This part of the simplifier may break the no-shadowing invariant
1235 f (...(\a -> e)...) (case y of (a,b) -> e')
1236 where f is strict in its second arg
1237 If we simplify the innermost one first we get (...(\a -> e)...)
1238 Simplifying the second arg makes us float the case out, so we end up with
1239 case y of (a,b) -> f (...(\a -> e)...) e'
1240 So the output does not have the no-shadowing invariant. However, there is
1241 no danger of getting name-capture, because when the first arg was simplified
1242 we used an in-scope set that at least mentioned all the variables free in its
1243 static environment, and that is enough.
1245 We can't just do innermost first, or we'd end up with a dual problem:
1246 case x of (a,b) -> f e (...(\a -> e')...)
1248 I spent hours trying to recover the no-shadowing invariant, but I just could
1249 not think of an elegant way to do it. The simplifier is already knee-deep in
1250 continuations. We have to keep the right in-scope set around; AND we have
1251 to get the effect that finding (error "foo") in a strict arg position will
1252 discard the entire application and replace it with (error "foo"). Getting
1253 all this at once is TOO HARD!
1256 %************************************************************************
1260 %************************************************************************
1263 tryRules :: SimplEnv -> [CoreRule]
1264 -> Id -> [OutExpr] -> SimplCont
1265 -> SimplM (Maybe (Arity, CoreExpr)) -- The arity is the number of
1266 -- args consumed by the rule
1267 tryRules env rules fn args call_cont
1271 = do { dflags <- getDOptsSmpl
1272 ; case activeRule dflags env of {
1273 Nothing -> return Nothing ; -- No rules apply
1275 case lookupRule act_fn (activeUnfInRule env) (getInScope env) fn args rules of {
1276 Nothing -> return Nothing ; -- No rule matches
1277 Just (rule, rule_rhs) ->
1279 do { tick (RuleFired (ru_name rule))
1280 ; (if dopt Opt_D_dump_rule_firings dflags then
1281 pprTrace "Rule fired" (vcat [
1282 text "Rule:" <+> ftext (ru_name rule),
1283 text "Before:" <+> ppr fn <+> sep (map pprParendExpr args),
1284 text "After: " <+> pprCoreExpr rule_rhs,
1285 text "Cont: " <+> ppr call_cont])
1288 return (Just (ruleArity rule, rule_rhs)) }}}}
1291 Note [Rules for recursive functions]
1292 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1293 You might think that we shouldn't apply rules for a loop breaker:
1294 doing so might give rise to an infinite loop, because a RULE is
1295 rather like an extra equation for the function:
1296 RULE: f (g x) y = x+y
1299 But it's too drastic to disable rules for loop breakers.
1300 Even the foldr/build rule would be disabled, because foldr
1301 is recursive, and hence a loop breaker:
1302 foldr k z (build g) = g k z
1303 So it's up to the programmer: rules can cause divergence
1306 %************************************************************************
1308 Rebuilding a cse expression
1310 %************************************************************************
1312 Note [Case elimination]
1313 ~~~~~~~~~~~~~~~~~~~~~~~
1314 The case-elimination transformation discards redundant case expressions.
1315 Start with a simple situation:
1317 case x# of ===> e[x#/y#]
1320 (when x#, y# are of primitive type, of course). We can't (in general)
1321 do this for algebraic cases, because we might turn bottom into
1324 The code in SimplUtils.prepareAlts has the effect of generalise this
1325 idea to look for a case where we're scrutinising a variable, and we
1326 know that only the default case can match. For example:
1330 DEFAULT -> ...(case x of
1334 Here the inner case is first trimmed to have only one alternative, the
1335 DEFAULT, after which it's an instance of the previous case. This
1336 really only shows up in eliminating error-checking code.
1338 We also make sure that we deal with this very common case:
1343 Here we are using the case as a strict let; if x is used only once
1344 then we want to inline it. We have to be careful that this doesn't
1345 make the program terminate when it would have diverged before, so we
1347 - e is already evaluated (it may so if e is a variable)
1348 - x is used strictly, or
1350 Lastly, the code in SimplUtils.mkCase combines identical RHSs. So
1352 case e of ===> case e of DEFAULT -> r
1356 Now again the case may be elminated by the CaseElim transformation.
1359 Further notes about case elimination
1360 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1361 Consider: test :: Integer -> IO ()
1364 Turns out that this compiles to:
1367 eta1 :: State# RealWorld ->
1368 case PrelNum.< eta PrelNum.zeroInteger of wild { __DEFAULT ->
1370 (PrelNum.jtos eta ($w[] @ Char))
1372 of wild1 { (# new_s, a4 #) -> PrelIO.lvl23 new_s }}
1374 Notice the strange '<' which has no effect at all. This is a funny one.
1375 It started like this:
1377 f x y = if x < 0 then jtos x
1378 else if y==0 then "" else jtos x
1380 At a particular call site we have (f v 1). So we inline to get
1382 if v < 0 then jtos x
1383 else if 1==0 then "" else jtos x
1385 Now simplify the 1==0 conditional:
1387 if v<0 then jtos v else jtos v
1389 Now common-up the two branches of the case:
1391 case (v<0) of DEFAULT -> jtos v
1393 Why don't we drop the case? Because it's strict in v. It's technically
1394 wrong to drop even unnecessary evaluations, and in practice they
1395 may be a result of 'seq' so we *definitely* don't want to drop those.
1396 I don't really know how to improve this situation.
1399 ---------------------------------------------------------
1400 -- Eliminate the case if possible
1402 rebuildCase, reallyRebuildCase
1404 -> OutExpr -- Scrutinee
1405 -> InId -- Case binder
1406 -> [InAlt] -- Alternatives (inceasing order)
1408 -> SimplM (SimplEnv, OutExpr)
1410 --------------------------------------------------
1411 -- 1. Eliminate the case if there's a known constructor
1412 --------------------------------------------------
1414 rebuildCase env scrut case_bndr alts cont
1415 | Lit lit <- scrut -- No need for same treatment as constructors
1416 -- because literals are inlined more vigorously
1417 = do { tick (KnownBranch case_bndr)
1418 ; case findAlt (LitAlt lit) alts of
1419 Nothing -> missingAlt env case_bndr alts cont
1420 Just (_, bs, rhs) -> simple_rhs bs rhs }
1422 | Just (con, ty_args, other_args) <- exprIsConApp_maybe (activeUnfInRule env) scrut
1423 -- Works when the scrutinee is a variable with a known unfolding
1424 -- as well as when it's an explicit constructor application
1425 = do { tick (KnownBranch case_bndr)
1426 ; case findAlt (DataAlt con) alts of
1427 Nothing -> missingAlt env case_bndr alts cont
1428 Just (DEFAULT, bs, rhs) -> simple_rhs bs rhs
1429 Just (_, bs, rhs) -> knownCon env scrut con ty_args other_args
1430 case_bndr bs rhs cont
1433 simple_rhs bs rhs = ASSERT( null bs )
1434 do { env' <- simplNonRecX env case_bndr scrut
1435 ; simplExprF env' rhs cont }
1438 --------------------------------------------------
1439 -- 2. Eliminate the case if scrutinee is evaluated
1440 --------------------------------------------------
1442 rebuildCase env scrut case_bndr [(_, bndrs, rhs)] cont
1443 -- See if we can get rid of the case altogether
1444 -- See Note [Case eliminiation]
1445 -- mkCase made sure that if all the alternatives are equal,
1446 -- then there is now only one (DEFAULT) rhs
1447 | all isDeadBinder bndrs -- bndrs are [InId]
1449 -- Check that the scrutinee can be let-bound instead of case-bound
1450 , exprOkForSpeculation scrut
1451 -- OK not to evaluate it
1452 -- This includes things like (==# a# b#)::Bool
1453 -- so that we simplify
1454 -- case ==# a# b# of { True -> x; False -> x }
1457 -- This particular example shows up in default methods for
1458 -- comparision operations (e.g. in (>=) for Int.Int32)
1459 || exprIsHNF scrut -- It's already evaluated
1460 || var_demanded_later scrut -- It'll be demanded later
1462 -- || not opt_SimplPedanticBottoms) -- Or we don't care!
1463 -- We used to allow improving termination by discarding cases, unless -fpedantic-bottoms was on,
1464 -- but that breaks badly for the dataToTag# primop, which relies on a case to evaluate
1465 -- its argument: case x of { y -> dataToTag# y }
1466 -- Here we must *not* discard the case, because dataToTag# just fetches the tag from
1467 -- the info pointer. So we'll be pedantic all the time, and see if that gives any
1469 -- Also we don't want to discard 'seq's
1470 = do { tick (CaseElim case_bndr)
1471 ; env' <- simplNonRecX env case_bndr scrut
1472 ; simplExprF env' rhs cont }
1474 -- The case binder is going to be evaluated later,
1475 -- and the scrutinee is a simple variable
1476 var_demanded_later (Var v) = isStrictDmd (idDemandInfo case_bndr)
1477 && not (isTickBoxOp v)
1478 -- ugly hack; covering this case is what
1479 -- exprOkForSpeculation was intended for.
1480 var_demanded_later _ = False
1482 --------------------------------------------------
1483 -- 3. Try seq rules; see Note [User-defined RULES for seq] in MkId
1484 --------------------------------------------------
1486 rebuildCase env scrut case_bndr alts@[(_, bndrs, rhs)] cont
1487 | all isDeadBinder (case_bndr : bndrs) -- So this is just 'seq'
1488 = do { let rhs' = substExpr env rhs
1489 out_args = [Type (substTy env (idType case_bndr)),
1490 Type (exprType rhs'), scrut, rhs']
1491 -- Lazily evaluated, so we don't do most of this
1493 ; rule_base <- getSimplRules
1494 ; mb_rule <- tryRules env (getRules rule_base seqId) seqId out_args cont
1496 Just (n_args, res) -> simplExprF (zapSubstEnv env)
1497 (mkApps res (drop n_args out_args))
1499 Nothing -> reallyRebuildCase env scrut case_bndr alts cont }
1501 rebuildCase env scrut case_bndr alts cont
1502 = reallyRebuildCase env scrut case_bndr alts cont
1504 --------------------------------------------------
1505 -- 3. Catch-all case
1506 --------------------------------------------------
1508 reallyRebuildCase env scrut case_bndr alts cont
1509 = do { -- Prepare the continuation;
1510 -- The new subst_env is in place
1511 (env', dup_cont, nodup_cont) <- prepareCaseCont env alts cont
1513 -- Simplify the alternatives
1514 ; (scrut', case_bndr', alts') <- simplAlts env' scrut case_bndr alts dup_cont
1516 -- Check for empty alternatives
1517 ; if null alts' then missingAlt env case_bndr alts cont
1519 { dflags <- getDOptsSmpl
1520 ; case_expr <- mkCase dflags scrut' case_bndr' alts'
1522 -- Notice that rebuild gets the in-scope set from env', not alt_env
1523 -- (which in any case is only build in simplAlts)
1524 -- The case binder *not* scope over the whole returned case-expression
1525 ; rebuild env' case_expr nodup_cont } }
1528 simplCaseBinder checks whether the scrutinee is a variable, v. If so,
1529 try to eliminate uses of v in the RHSs in favour of case_bndr; that
1530 way, there's a chance that v will now only be used once, and hence
1533 Historical note: we use to do the "case binder swap" in the Simplifier
1534 so there were additional complications if the scrutinee was a variable.
1535 Now the binder-swap stuff is done in the occurrence analyer; see
1536 OccurAnal Note [Binder swap].
1540 If the case binder is not dead, then neither are the pattern bound
1542 case <any> of x { (a,b) ->
1543 case x of { (p,q) -> p } }
1544 Here (a,b) both look dead, but come alive after the inner case is eliminated.
1545 The point is that we bring into the envt a binding
1547 after the outer case, and that makes (a,b) alive. At least we do unless
1548 the case binder is guaranteed dead.
1550 In practice, the scrutinee is almost always a variable, so we pretty
1551 much always zap the OccInfo of the binders. It doesn't matter much though.
1556 Consider case (v `cast` co) of x { I# y ->
1557 ... (case (v `cast` co) of {...}) ...
1558 We'd like to eliminate the inner case. We can get this neatly by
1559 arranging that inside the outer case we add the unfolding
1560 v |-> x `cast` (sym co)
1561 to v. Then we should inline v at the inner case, cancel the casts, and away we go
1563 Note [Improving seq]
1566 type family F :: * -> *
1567 type instance F Int = Int
1569 ... case e of x { DEFAULT -> rhs } ...
1571 where x::F Int. Then we'd like to rewrite (F Int) to Int, getting
1573 case e `cast` co of x'::Int
1574 I# x# -> let x = x' `cast` sym co
1577 so that 'rhs' can take advantage of the form of x'.
1579 Notice that Note [Case of cast] may then apply to the result.
1581 Nota Bene: We only do the [Improving seq] transformation if the
1582 case binder 'x' is actually used in the rhs; that is, if the case
1583 is *not* a *pure* seq.
1584 a) There is no point in adding the cast to a pure seq.
1585 b) There is a good reason not to: doing so would interfere
1586 with seq rules (Note [Built-in RULES for seq] in MkId).
1587 In particular, this [Improving seq] thing *adds* a cast
1588 while [Built-in RULES for seq] *removes* one, so they
1591 You might worry about
1592 case v of x { __DEFAULT ->
1593 ... case (v `cast` co) of y { I# -> ... }}
1594 This is a pure seq (since x is unused), so [Improving seq] won't happen.
1595 But it's ok: the simplifier will replace 'v' by 'x' in the rhs to get
1596 case v of x { __DEFAULT ->
1597 ... case (x `cast` co) of y { I# -> ... }}
1598 Now the outer case is not a pure seq, so [Improving seq] will happen,
1599 and then the inner case will disappear.
1601 The need for [Improving seq] showed up in Roman's experiments. Example:
1602 foo :: F Int -> Int -> Int
1603 foo t n = t `seq` bar n
1606 bar n = bar (n - case t of TI i -> i)
1607 Here we'd like to avoid repeated evaluating t inside the loop, by
1608 taking advantage of the `seq`.
1610 At one point I did transformation in LiberateCase, but it's more
1611 robust here. (Otherwise, there's a danger that we'll simply drop the
1612 'seq' altogether, before LiberateCase gets to see it.)
1615 simplAlts :: SimplEnv
1617 -> InId -- Case binder
1618 -> [InAlt] -- Non-empty
1620 -> SimplM (OutExpr, OutId, [OutAlt]) -- Includes the continuation
1621 -- Like simplExpr, this just returns the simplified alternatives;
1622 -- it does not return an environment
1624 simplAlts env scrut case_bndr alts cont'
1625 = -- pprTrace "simplAlts" (ppr alts $$ ppr (seIdSubst env)) $
1626 do { let env0 = zapFloats env
1628 ; (env1, case_bndr1) <- simplBinder env0 case_bndr
1630 ; fam_envs <- getFamEnvs
1631 ; (alt_env', scrut', case_bndr') <- improveSeq fam_envs env1 scrut
1632 case_bndr case_bndr1 alts
1634 ; (imposs_deflt_cons, in_alts) <- prepareAlts scrut' case_bndr' alts
1636 ; alts' <- mapM (simplAlt alt_env' imposs_deflt_cons case_bndr' cont') in_alts
1637 ; return (scrut', case_bndr', alts') }
1640 ------------------------------------
1641 improveSeq :: (FamInstEnv, FamInstEnv) -> SimplEnv
1642 -> OutExpr -> InId -> OutId -> [InAlt]
1643 -> SimplM (SimplEnv, OutExpr, OutId)
1644 -- Note [Improving seq]
1645 improveSeq fam_envs env scrut case_bndr case_bndr1 [(DEFAULT,_,_)]
1646 | not (isDeadBinder case_bndr) -- Not a pure seq! See the Note!
1647 , Just (co, ty2) <- topNormaliseType fam_envs (idType case_bndr1)
1648 = do { case_bndr2 <- newId (fsLit "nt") ty2
1649 ; let rhs = DoneEx (Var case_bndr2 `Cast` mkSymCoercion co)
1650 env2 = extendIdSubst env case_bndr rhs
1651 ; return (env2, scrut `Cast` co, case_bndr2) }
1653 improveSeq _ env scrut _ case_bndr1 _
1654 = return (env, scrut, case_bndr1)
1657 ------------------------------------
1658 simplAlt :: SimplEnv
1659 -> [AltCon] -- These constructors can't be present when
1660 -- matching the DEFAULT alternative
1661 -> OutId -- The case binder
1666 simplAlt env imposs_deflt_cons case_bndr' cont' (DEFAULT, bndrs, rhs)
1667 = ASSERT( null bndrs )
1668 do { let env' = addBinderOtherCon env case_bndr' imposs_deflt_cons
1669 -- Record the constructors that the case-binder *can't* be.
1670 ; rhs' <- simplExprC env' rhs cont'
1671 ; return (DEFAULT, [], rhs') }
1673 simplAlt env _ case_bndr' cont' (LitAlt lit, bndrs, rhs)
1674 = ASSERT( null bndrs )
1675 do { let env' = addBinderUnfolding env case_bndr' (Lit lit)
1676 ; rhs' <- simplExprC env' rhs cont'
1677 ; return (LitAlt lit, [], rhs') }
1679 simplAlt env _ case_bndr' cont' (DataAlt con, vs, rhs)
1680 = do { -- Deal with the pattern-bound variables
1681 -- Mark the ones that are in ! positions in the
1682 -- data constructor as certainly-evaluated.
1683 -- NB: simplLamBinders preserves this eval info
1684 let vs_with_evals = add_evals (dataConRepStrictness con)
1685 ; (env', vs') <- simplLamBndrs env vs_with_evals
1687 -- Bind the case-binder to (con args)
1688 ; let inst_tys' = tyConAppArgs (idType case_bndr')
1689 con_args = map Type inst_tys' ++ varsToCoreExprs vs'
1690 env'' = addBinderUnfolding env' case_bndr'
1691 (mkConApp con con_args)
1693 ; rhs' <- simplExprC env'' rhs cont'
1694 ; return (DataAlt con, vs', rhs') }
1696 -- add_evals records the evaluated-ness of the bound variables of
1697 -- a case pattern. This is *important*. Consider
1698 -- data T = T !Int !Int
1700 -- case x of { T a b -> T (a+1) b }
1702 -- We really must record that b is already evaluated so that we don't
1703 -- go and re-evaluate it when constructing the result.
1704 -- See Note [Data-con worker strictness] in MkId.lhs
1709 go (v:vs') strs | isTyVar v = v : go vs' strs
1710 go (v:vs') (str:strs)
1711 | isMarkedStrict str = evald_v : go vs' strs
1712 | otherwise = zapped_v : go vs' strs
1714 zapped_v = zap_occ_info v
1715 evald_v = zapped_v `setIdUnfolding` evaldUnfolding
1716 go _ _ = pprPanic "cat_evals" (ppr con $$ ppr vs $$ ppr the_strs)
1718 -- See Note [zapOccInfo]
1719 -- zap_occ_info: if the case binder is alive, then we add the unfolding
1721 -- to the envt; so vs are now very much alive
1722 -- Note [Aug06] I can't see why this actually matters, but it's neater
1723 -- case e of t { (a,b) -> ...(case t of (p,q) -> p)... }
1724 -- ==> case e of t { (a,b) -> ...(a)... }
1725 -- Look, Ma, a is alive now.
1726 zap_occ_info = zapCasePatIdOcc case_bndr'
1728 addBinderUnfolding :: SimplEnv -> Id -> CoreExpr -> SimplEnv
1729 addBinderUnfolding env bndr rhs
1730 = modifyInScope env (bndr `setIdUnfolding` mkUnfolding False False rhs)
1732 addBinderOtherCon :: SimplEnv -> Id -> [AltCon] -> SimplEnv
1733 addBinderOtherCon env bndr cons
1734 = modifyInScope env (bndr `setIdUnfolding` mkOtherCon cons)
1736 zapCasePatIdOcc :: Id -> Id -> Id
1737 -- Consider case e of b { (a,b) -> ... }
1738 -- Then if we bind b to (a,b) in "...", and b is not dead,
1739 -- then we must zap the deadness info on a,b
1740 zapCasePatIdOcc case_bndr
1741 | isDeadBinder case_bndr = \ pat_id -> pat_id
1742 | otherwise = \ pat_id -> zapIdOccInfo pat_id
1746 %************************************************************************
1748 \subsection{Known constructor}
1750 %************************************************************************
1752 We are a bit careful with occurrence info. Here's an example
1754 (\x* -> case x of (a*, b) -> f a) (h v, e)
1756 where the * means "occurs once". This effectively becomes
1757 case (h v, e) of (a*, b) -> f a)
1759 let a* = h v; b = e in f a
1763 All this should happen in one sweep.
1766 knownCon :: SimplEnv
1767 -> OutExpr -- The scrutinee
1768 -> DataCon -> [OutType] -> [OutExpr] -- The scrutinee (in pieces)
1769 -> InId -> [InBndr] -> InExpr -- The alternative
1771 -> SimplM (SimplEnv, OutExpr)
1773 knownCon env scrut dc dc_ty_args dc_args bndr bs rhs cont
1774 = do { env' <- bind_args env bs dc_args
1776 -- It's useful to bind bndr to scrut, rather than to a fresh
1777 -- binding x = Con arg1 .. argn
1778 -- because very often the scrut is a variable, so we avoid
1779 -- creating, and then subsequently eliminating, a let-binding
1780 -- BUT, if scrut is a not a variable, we must be careful
1781 -- about duplicating the arg redexes; in that case, make
1782 -- a new con-app from the args
1783 bndr_rhs | exprIsTrivial scrut = scrut
1784 | otherwise = con_app
1785 con_app = Var (dataConWorkId dc)
1786 `mkTyApps` dc_ty_args
1787 `mkApps` [substExpr env' (varToCoreExpr b) | b <- bs]
1788 -- dc_ty_args are aready OutTypes, but bs are InBndrs
1790 ; env'' <- simplNonRecX env' bndr bndr_rhs
1791 ; simplExprF env'' rhs cont }
1793 zap_occ = zapCasePatIdOcc bndr -- bndr is an InId
1796 bind_args env' [] _ = return env'
1798 bind_args env' (b:bs') (Type ty : args)
1799 = ASSERT( isTyVar b )
1800 bind_args (extendTvSubst env' b ty) bs' args
1802 bind_args env' (b:bs') (arg : args)
1804 do { let b' = zap_occ b
1805 -- Note that the binder might be "dead", because it doesn't
1806 -- occur in the RHS; and simplNonRecX may therefore discard
1807 -- it via postInlineUnconditionally.
1808 -- Nevertheless we must keep it if the case-binder is alive,
1809 -- because it may be used in the con_app. See Note [zapOccInfo]
1810 ; env'' <- simplNonRecX env' b' arg
1811 ; bind_args env'' bs' args }
1814 pprPanic "bind_args" $ ppr dc $$ ppr bs $$ ppr dc_args $$
1815 text "scrut:" <+> ppr scrut
1818 missingAlt :: SimplEnv -> Id -> [InAlt] -> SimplCont -> SimplM (SimplEnv, OutExpr)
1819 -- This isn't strictly an error, although it is unusual.
1820 -- It's possible that the simplifer might "see" that
1821 -- an inner case has no accessible alternatives before
1822 -- it "sees" that the entire branch of an outer case is
1823 -- inaccessible. So we simply put an error case here instead.
1824 missingAlt env case_bndr alts cont
1825 = WARN( True, ptext (sLit "missingAlt") <+> ppr case_bndr )
1826 return (env, mkImpossibleExpr res_ty)
1828 res_ty = contResultType env (substTy env (coreAltsType alts)) cont
1832 %************************************************************************
1834 \subsection{Duplicating continuations}
1836 %************************************************************************
1839 prepareCaseCont :: SimplEnv
1840 -> [InAlt] -> SimplCont
1841 -> SimplM (SimplEnv, SimplCont,SimplCont)
1842 -- Return a duplicatable continuation, a non-duplicable part
1843 -- plus some extra bindings (that scope over the entire
1846 -- No need to make it duplicatable if there's only one alternative
1847 prepareCaseCont env [_] cont = return (env, cont, mkBoringStop)
1848 prepareCaseCont env _ cont = mkDupableCont env cont
1852 mkDupableCont :: SimplEnv -> SimplCont
1853 -> SimplM (SimplEnv, SimplCont, SimplCont)
1855 mkDupableCont env cont
1856 | contIsDupable cont
1857 = return (env, cont, mkBoringStop)
1859 mkDupableCont _ (Stop {}) = panic "mkDupableCont" -- Handled by previous eqn
1861 mkDupableCont env (CoerceIt ty cont)
1862 = do { (env', dup, nodup) <- mkDupableCont env cont
1863 ; return (env', CoerceIt ty dup, nodup) }
1865 mkDupableCont env cont@(StrictBind {})
1866 = return (env, mkBoringStop, cont)
1867 -- See Note [Duplicating StrictBind]
1869 mkDupableCont env (StrictArg info cci cont)
1870 -- See Note [Duplicating StrictArg]
1871 = do { (env', dup, nodup) <- mkDupableCont env cont
1872 ; (env'', args') <- mapAccumLM makeTrivial env' (ai_args info)
1873 ; return (env'', StrictArg (info { ai_args = args' }) cci dup, nodup) }
1875 mkDupableCont env (ApplyTo _ arg se cont)
1876 = -- e.g. [...hole...] (...arg...)
1878 -- let a = ...arg...
1879 -- in [...hole...] a
1880 do { (env', dup_cont, nodup_cont) <- mkDupableCont env cont
1881 ; arg' <- simplExpr (se `setInScope` env') arg
1882 ; (env'', arg'') <- makeTrivial env' arg'
1883 ; let app_cont = ApplyTo OkToDup arg'' (zapSubstEnv env'') dup_cont
1884 ; return (env'', app_cont, nodup_cont) }
1886 mkDupableCont env cont@(Select _ case_bndr [(_, bs, _rhs)] _ _)
1887 -- See Note [Single-alternative case]
1888 -- | not (exprIsDupable rhs && contIsDupable case_cont)
1889 -- | not (isDeadBinder case_bndr)
1890 | all isDeadBinder bs -- InIds
1891 && not (isUnLiftedType (idType case_bndr))
1892 -- Note [Single-alternative-unlifted]
1893 = return (env, mkBoringStop, cont)
1895 mkDupableCont env (Select _ case_bndr alts se cont)
1896 = -- e.g. (case [...hole...] of { pi -> ei })
1898 -- let ji = \xij -> ei
1899 -- in case [...hole...] of { pi -> ji xij }
1900 do { tick (CaseOfCase case_bndr)
1901 ; (env', dup_cont, nodup_cont) <- mkDupableCont env cont
1902 -- NB: call mkDupableCont here, *not* prepareCaseCont
1903 -- We must make a duplicable continuation, whereas prepareCaseCont
1904 -- doesn't when there is a single case branch
1906 ; let alt_env = se `setInScope` env'
1907 ; (alt_env', case_bndr') <- simplBinder alt_env case_bndr
1908 ; alts' <- mapM (simplAlt alt_env' [] case_bndr' dup_cont) alts
1909 -- Safe to say that there are no handled-cons for the DEFAULT case
1910 -- NB: simplBinder does not zap deadness occ-info, so
1911 -- a dead case_bndr' will still advertise its deadness
1912 -- This is really important because in
1913 -- case e of b { (# p,q #) -> ... }
1914 -- b is always dead, and indeed we are not allowed to bind b to (# p,q #),
1915 -- which might happen if e was an explicit unboxed pair and b wasn't marked dead.
1916 -- In the new alts we build, we have the new case binder, so it must retain
1918 -- NB: we don't use alt_env further; it has the substEnv for
1919 -- the alternatives, and we don't want that
1921 ; (env'', alts'') <- mkDupableAlts env' case_bndr' alts'
1922 ; return (env'', -- Note [Duplicated env]
1923 Select OkToDup case_bndr' alts'' (zapSubstEnv env'') mkBoringStop,
1927 mkDupableAlts :: SimplEnv -> OutId -> [InAlt]
1928 -> SimplM (SimplEnv, [InAlt])
1929 -- Absorbs the continuation into the new alternatives
1931 mkDupableAlts env case_bndr' the_alts
1934 go env0 [] = return (env0, [])
1936 = do { (env1, alt') <- mkDupableAlt env0 case_bndr' alt
1937 ; (env2, alts') <- go env1 alts
1938 ; return (env2, alt' : alts' ) }
1940 mkDupableAlt :: SimplEnv -> OutId -> (AltCon, [CoreBndr], CoreExpr)
1941 -> SimplM (SimplEnv, (AltCon, [CoreBndr], CoreExpr))
1942 mkDupableAlt env case_bndr (con, bndrs', rhs')
1943 | exprIsDupable rhs' -- Note [Small alternative rhs]
1944 = return (env, (con, bndrs', rhs'))
1946 = do { let rhs_ty' = exprType rhs'
1947 scrut_ty = idType case_bndr
1950 DEFAULT -> case_bndr
1951 DataAlt dc -> setIdUnfolding case_bndr unf
1953 -- See Note [Case binders and join points]
1954 unf = mkInlineRule needSaturated rhs 0
1955 rhs = mkConApp dc (map Type (tyConAppArgs scrut_ty)
1956 ++ varsToCoreExprs bndrs')
1958 LitAlt {} -> WARN( True, ptext (sLit "mkDupableAlt")
1959 <+> ppr case_bndr <+> ppr con )
1961 -- The case binder is alive but trivial, so why has
1962 -- it not been substituted away?
1964 used_bndrs' | isDeadBinder case_bndr = filter abstract_over bndrs'
1965 | otherwise = bndrs' ++ [case_bndr_w_unf]
1968 | isTyVar bndr = True -- Abstract over all type variables just in case
1969 | otherwise = not (isDeadBinder bndr)
1970 -- The deadness info on the new Ids is preserved by simplBinders
1972 ; (final_bndrs', final_args) -- Note [Join point abstraction]
1973 <- if (any isId used_bndrs')
1974 then return (used_bndrs', varsToCoreExprs used_bndrs')
1975 else do { rw_id <- newId (fsLit "w") realWorldStatePrimTy
1976 ; return ([rw_id], [Var realWorldPrimId]) }
1978 ; join_bndr <- newId (fsLit "$j") (mkPiTypes final_bndrs' rhs_ty')
1979 -- Note [Funky mkPiTypes]
1981 ; let -- We make the lambdas into one-shot-lambdas. The
1982 -- join point is sure to be applied at most once, and doing so
1983 -- prevents the body of the join point being floated out by
1984 -- the full laziness pass
1985 really_final_bndrs = map one_shot final_bndrs'
1986 one_shot v | isId v = setOneShotLambda v
1988 join_rhs = mkLams really_final_bndrs rhs'
1989 join_call = mkApps (Var join_bndr) final_args
1991 ; env' <- addPolyBind NotTopLevel env (NonRec join_bndr join_rhs)
1992 ; return (env', (con, bndrs', join_call)) }
1993 -- See Note [Duplicated env]
1996 Note [Case binders and join points]
1997 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1999 case (case .. ) of c {
2002 If we make a join point with c but not c# we get
2003 $j = \c -> ....c....
2005 But if later inlining scrutines the c, thus
2007 $j = \c -> ... case c of { I# y -> ... } ...
2009 we won't see that 'c' has already been scrutinised. This actually
2010 happens in the 'tabulate' function in wave4main, and makes a significant
2011 difference to allocation.
2013 An alternative plan is this:
2015 $j = \c# -> let c = I# c# in ...c....
2017 but that is bad if 'c' is *not* later scrutinised.
2019 So instead we do both: we pass 'c' and 'c#' , and record in c's inlining
2020 that it's really I# c#, thus
2022 $j = \c# -> \c[=I# c#] -> ...c....
2024 Absence analysis may later discard 'c'.
2027 Note [Duplicated env]
2028 ~~~~~~~~~~~~~~~~~~~~~
2029 Some of the alternatives are simplified, but have not been turned into a join point
2030 So they *must* have an zapped subst-env. So we can't use completeNonRecX to
2031 bind the join point, because it might to do PostInlineUnconditionally, and
2032 we'd lose that when zapping the subst-env. We could have a per-alt subst-env,
2033 but zapping it (as we do in mkDupableCont, the Select case) is safe, and
2034 at worst delays the join-point inlining.
2036 Note [Small alternative rhs]
2037 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2038 It is worth checking for a small RHS because otherwise we
2039 get extra let bindings that may cause an extra iteration of the simplifier to
2040 inline back in place. Quite often the rhs is just a variable or constructor.
2041 The Ord instance of Maybe in PrelMaybe.lhs, for example, took several extra
2042 iterations because the version with the let bindings looked big, and so wasn't
2043 inlined, but after the join points had been inlined it looked smaller, and so
2046 NB: we have to check the size of rhs', not rhs.
2047 Duplicating a small InAlt might invalidate occurrence information
2048 However, if it *is* dupable, we return the *un* simplified alternative,
2049 because otherwise we'd need to pair it up with an empty subst-env....
2050 but we only have one env shared between all the alts.
2051 (Remember we must zap the subst-env before re-simplifying something).
2052 Rather than do this we simply agree to re-simplify the original (small) thing later.
2054 Note [Funky mkPiTypes]
2055 ~~~~~~~~~~~~~~~~~~~~~~
2056 Notice the funky mkPiTypes. If the contructor has existentials
2057 it's possible that the join point will be abstracted over
2058 type varaibles as well as term variables.
2059 Example: Suppose we have
2060 data T = forall t. C [t]
2062 case (case e of ...) of
2064 We get the join point
2065 let j :: forall t. [t] -> ...
2066 j = /\t \xs::[t] -> rhs
2068 case (case e of ...) of
2069 C t xs::[t] -> j t xs
2071 Note [Join point abstaction]
2072 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2073 If we try to lift a primitive-typed something out
2074 for let-binding-purposes, we will *caseify* it (!),
2075 with potentially-disastrous strictness results. So
2076 instead we turn it into a function: \v -> e
2077 where v::State# RealWorld#. The value passed to this function
2078 is realworld#, which generates (almost) no code.
2080 There's a slight infelicity here: we pass the overall
2081 case_bndr to all the join points if it's used in *any* RHS,
2082 because we don't know its usage in each RHS separately
2084 We used to say "&& isUnLiftedType rhs_ty'" here, but now
2085 we make the join point into a function whenever used_bndrs'
2086 is empty. This makes the join-point more CPR friendly.
2087 Consider: let j = if .. then I# 3 else I# 4
2088 in case .. of { A -> j; B -> j; C -> ... }
2090 Now CPR doesn't w/w j because it's a thunk, so
2091 that means that the enclosing function can't w/w either,
2092 which is a lose. Here's the example that happened in practice:
2093 kgmod :: Int -> Int -> Int
2094 kgmod x y = if x > 0 && y < 0 || x < 0 && y > 0
2098 I have seen a case alternative like this:
2100 It's a bit silly to add the realWorld dummy arg in this case, making
2103 (the \v alone is enough to make CPR happy) but I think it's rare
2105 Note [Duplicating StrictArg]
2106 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2107 The original plan had (where E is a big argument)
2109 ==> let $j = \a -> f E a
2112 But this is terrible! Here's an example:
2113 && E (case x of { T -> F; F -> T })
2114 Now, && is strict so we end up simplifying the case with
2115 an ArgOf continuation. If we let-bind it, we get
2116 let $j = \v -> && E v
2117 in simplExpr (case x of { T -> F; F -> T })
2119 And after simplifying more we get
2120 let $j = \v -> && E v
2121 in case x of { T -> $j F; F -> $j T }
2122 Which is a Very Bad Thing
2124 What we do now is this
2128 Now if the thing in the hole is a case expression (which is when
2129 we'll call mkDupableCont), we'll push the function call into the
2130 branches, which is what we want. Now RULES for f may fire, and
2131 call-pattern specialisation. Here's an example from Trac #3116
2134 _ -> Chunk p fpc (o+1) (l-1) bs')
2135 If we can push the call for 'go' inside the case, we get
2136 call-pattern specialisation for 'go', which is *crucial* for
2139 Here is the (&&) example:
2140 && E (case x of { T -> F; F -> T })
2142 case x of { T -> && a F; F -> && a T }
2146 * Arguments to f *after* the strict one are handled by
2147 the ApplyTo case of mkDupableCont. Eg
2150 * We can only do the let-binding of E because the function
2151 part of a StrictArg continuation is an explicit syntax
2152 tree. In earlier versions we represented it as a function
2153 (CoreExpr -> CoreEpxr) which we couldn't take apart.
2155 Do *not* duplicate StrictBind and StritArg continuations. We gain
2156 nothing by propagating them into the expressions, and we do lose a
2159 The desire not to duplicate is the entire reason that
2160 mkDupableCont returns a pair of continuations.
2162 Note [Duplicating StrictBind]
2163 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2164 Unlike StrictArg, there doesn't seem anything to gain from
2165 duplicating a StrictBind continuation, so we don't.
2167 The desire not to duplicate is the entire reason that
2168 mkDupableCont returns a pair of continuations.
2171 Note [Single-alternative cases]
2172 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2173 This case is just like the ArgOf case. Here's an example:
2177 case (case x of I# x' ->
2179 True -> I# (negate# x')
2180 False -> I# x') of y {
2182 Because the (case x) has only one alternative, we'll transform to
2184 case (case x' <# 0# of
2185 True -> I# (negate# x')
2186 False -> I# x') of y {
2188 But now we do *NOT* want to make a join point etc, giving
2190 let $j = \y -> MkT y
2192 True -> $j (I# (negate# x'))
2194 In this case the $j will inline again, but suppose there was a big
2195 strict computation enclosing the orginal call to MkT. Then, it won't
2196 "see" the MkT any more, because it's big and won't get duplicated.
2197 And, what is worse, nothing was gained by the case-of-case transform.
2199 When should use this case of mkDupableCont?
2200 However, matching on *any* single-alternative case is a *disaster*;
2201 e.g. case (case ....) of (a,b) -> (# a,b #)
2202 We must push the outer case into the inner one!
2205 * Match [(DEFAULT,_,_)], but in the common case of Int,
2206 the alternative-filling-in code turned the outer case into
2207 case (...) of y { I# _ -> MkT y }
2209 * Match on single alternative plus (not (isDeadBinder case_bndr))
2210 Rationale: pushing the case inwards won't eliminate the construction.
2211 But there's a risk of
2212 case (...) of y { (a,b) -> let z=(a,b) in ... }
2213 Now y looks dead, but it'll come alive again. Still, this
2214 seems like the best option at the moment.
2216 * Match on single alternative plus (all (isDeadBinder bndrs))
2217 Rationale: this is essentially seq.
2219 * Match when the rhs is *not* duplicable, and hence would lead to a
2220 join point. This catches the disaster-case above. We can test
2221 the *un-simplified* rhs, which is fine. It might get bigger or
2222 smaller after simplification; if it gets smaller, this case might
2223 fire next time round. NB also that we must test contIsDupable
2224 case_cont *btoo, because case_cont might be big!
2226 HOWEVER: I found that this version doesn't work well, because
2227 we can get let x = case (...) of { small } in ...case x...
2228 When x is inlined into its full context, we find that it was a bad
2229 idea to have pushed the outer case inside the (...) case.
2231 Note [Single-alternative-unlifted]
2232 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2233 Here's another single-alternative where we really want to do case-of-case:
2241 case y_s6X of tpl_s7m {
2242 M1.Mk1 ipv_s70 -> ipv_s70;
2243 M1.Mk2 ipv_s72 -> ipv_s72;
2249 case x_s74 of tpl_s7n {
2250 M1.Mk1 ipv_s77 -> ipv_s77;
2251 M1.Mk2 ipv_s79 -> ipv_s79;
2255 { __DEFAULT -> ==# [wild1_s7b wild_s7c];
2259 So the outer case is doing *nothing at all*, other than serving as a
2260 join-point. In this case we really want to do case-of-case and decide
2261 whether to use a real join point or just duplicate the continuation.
2263 Hence: check whether the case binder's type is unlifted, because then
2264 the outer case is *not* a seq.