2 % (c) The AQUA Project, Glasgow University, 1993-1998
4 \section[Simplify]{The main module of the simplifier}
7 module Simplify ( simplTopBinds, simplExpr ) where
9 #include "HsVersions.h"
11 import CmdLineOpts ( dopt, DynFlag(Opt_D_dump_inlinings),
15 import SimplUtils ( mkCase, mkLam, newId, prepareAlts,
16 simplBinder, simplBinders, simplLamBndrs, simplRecBndrs, simplLetBndr,
17 SimplCont(..), DupFlag(..), LetRhsFlag(..),
18 mkStop, mkBoringStop, pushContArgs,
19 contResultType, countArgs, contIsDupable, contIsRhsOrArg,
20 getContArgs, interestingCallContext, interestingArg, isStrictType
22 import Var ( mustHaveLocalBinding )
24 import Id ( Id, idType, idInfo, idArity, isDataConWorkId,
25 setIdUnfolding, isDeadBinder,
26 idNewDemandInfo, setIdInfo,
27 setIdOccInfo, zapLamIdInfo, setOneShotLambda,
29 import OccName ( encodeFS )
30 import IdInfo ( OccInfo(..), isLoopBreaker,
35 import NewDemand ( isStrictDmd )
36 import DataCon ( dataConNumInstArgs, dataConRepStrictness )
38 import PprCore ( pprParendExpr, pprCoreExpr )
39 import CoreUnfold ( mkOtherCon, mkUnfolding, callSiteInline )
40 import CoreUtils ( exprIsDupable, exprIsTrivial, needsCaseBinding,
41 exprIsConApp_maybe, mkPiTypes, findAlt,
42 exprType, exprIsValue,
43 exprOkForSpeculation, exprArity,
44 mkCoerce, mkCoerce2, mkSCC, mkInlineMe, mkAltExpr, applyTypeToArg
46 import Rules ( lookupRule )
47 import BasicTypes ( isMarkedStrict )
48 import CostCentre ( currentCCS )
49 import Type ( isUnLiftedType, seqType, tyConAppArgs, funArgTy,
50 splitFunTy_maybe, splitFunTy, eqType
52 import Subst ( mkSubst, substTy, substExpr,
53 isInScope, lookupIdSubst, simplIdInfo
55 import TysPrim ( realWorldStatePrimTy )
56 import PrelInfo ( realWorldPrimId )
57 import BasicTypes ( TopLevelFlag(..), isTopLevel,
61 import Maybe ( Maybe )
63 import Util ( notNull )
67 The guts of the simplifier is in this module, but the driver loop for
68 the simplifier is in SimplCore.lhs.
71 -----------------------------------------
72 *** IMPORTANT NOTE ***
73 -----------------------------------------
74 The simplifier used to guarantee that the output had no shadowing, but
75 it does not do so any more. (Actually, it never did!) The reason is
76 documented with simplifyArgs.
79 -----------------------------------------
80 *** IMPORTANT NOTE ***
81 -----------------------------------------
82 Many parts of the simplifier return a bunch of "floats" as well as an
83 expression. This is wrapped as a datatype SimplUtils.FloatsWith.
85 All "floats" are let-binds, not case-binds, but some non-rec lets may
86 be unlifted (with RHS ok-for-speculation).
90 -----------------------------------------
91 ORGANISATION OF FUNCTIONS
92 -----------------------------------------
94 - simplify all top-level binders
95 - for NonRec, call simplRecOrTopPair
96 - for Rec, call simplRecBind
99 ------------------------------
100 simplExpr (applied lambda) ==> simplNonRecBind
101 simplExpr (Let (NonRec ...) ..) ==> simplNonRecBind
102 simplExpr (Let (Rec ...) ..) ==> simplify binders; simplRecBind
104 ------------------------------
105 simplRecBind [binders already simplfied]
106 - use simplRecOrTopPair on each pair in turn
108 simplRecOrTopPair [binder already simplified]
109 Used for: recursive bindings (top level and nested)
110 top-level non-recursive bindings
112 - check for PreInlineUnconditionally
116 Used for: non-top-level non-recursive bindings
117 beta reductions (which amount to the same thing)
118 Because it can deal with strict arts, it takes a
119 "thing-inside" and returns an expression
121 - check for PreInlineUnconditionally
122 - simplify binder, including its IdInfo
131 simplNonRecX: [given a *simplified* RHS, but an *unsimplified* binder]
132 Used for: binding case-binder and constr args in a known-constructor case
133 - check for PreInLineUnconditionally
137 ------------------------------
138 simplLazyBind: [binder already simplified, RHS not]
139 Used for: recursive bindings (top level and nested)
140 top-level non-recursive bindings
141 non-top-level, but *lazy* non-recursive bindings
142 [must not be strict or unboxed]
143 Returns floats + an augmented environment, not an expression
144 - substituteIdInfo and add result to in-scope
145 [so that rules are available in rec rhs]
148 - float if exposes constructor or PAP
152 completeNonRecX: [binder and rhs both simplified]
153 - if the the thing needs case binding (unlifted and not ok-for-spec)
159 completeLazyBind: [given a simplified RHS]
160 [used for both rec and non-rec bindings, top level and not]
161 - try PostInlineUnconditionally
162 - add unfolding [this is the only place we add an unfolding]
167 Right hand sides and arguments
168 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
169 In many ways we want to treat
170 (a) the right hand side of a let(rec), and
171 (b) a function argument
172 in the same way. But not always! In particular, we would
173 like to leave these arguments exactly as they are, so they
174 will match a RULE more easily.
179 It's harder to make the rule match if we ANF-ise the constructor,
180 or eta-expand the PAP:
182 f (let { a = g x; b = h x } in (a,b))
185 On the other hand if we see the let-defns
190 then we *do* want to ANF-ise and eta-expand, so that p and q
191 can be safely inlined.
193 Even floating lets out is a bit dubious. For let RHS's we float lets
194 out if that exposes a value, so that the value can be inlined more vigorously.
197 r = let x = e in (x,x)
199 Here, if we float the let out we'll expose a nice constructor. We did experiments
200 that showed this to be a generally good thing. But it was a bad thing to float
201 lets out unconditionally, because that meant they got allocated more often.
203 For function arguments, there's less reason to expose a constructor (it won't
204 get inlined). Just possibly it might make a rule match, but I'm pretty skeptical.
205 So for the moment we don't float lets out of function arguments either.
210 For eta expansion, we want to catch things like
212 case e of (a,b) -> \x -> case a of (p,q) -> \y -> r
214 If the \x was on the RHS of a let, we'd eta expand to bring the two
215 lambdas together. And in general that's a good thing to do. Perhaps
216 we should eta expand wherever we find a (value) lambda? Then the eta
217 expansion at a let RHS can concentrate solely on the PAP case.
220 %************************************************************************
222 \subsection{Bindings}
224 %************************************************************************
227 simplTopBinds :: SimplEnv -> [InBind] -> SimplM [OutBind]
229 simplTopBinds env binds
230 = -- Put all the top-level binders into scope at the start
231 -- so that if a transformation rule has unexpectedly brought
232 -- anything into scope, then we don't get a complaint about that.
233 -- It's rather as if the top-level binders were imported.
234 simplRecBndrs env (bindersOfBinds binds) `thenSmpl` \ (env, bndrs') ->
235 simpl_binds env binds bndrs' `thenSmpl` \ (floats, _) ->
236 freeTick SimplifierDone `thenSmpl_`
237 returnSmpl (floatBinds floats)
239 -- We need to track the zapped top-level binders, because
240 -- they should have their fragile IdInfo zapped (notably occurrence info)
241 -- That's why we run down binds and bndrs' simultaneously.
242 simpl_binds :: SimplEnv -> [InBind] -> [OutId] -> SimplM (FloatsWith ())
243 simpl_binds env [] bs = ASSERT( null bs ) returnSmpl (emptyFloats env, ())
244 simpl_binds env (bind:binds) bs = simpl_bind env bind bs `thenSmpl` \ (floats,env) ->
245 addFloats env floats $ \env ->
246 simpl_binds env binds (drop_bs bind bs)
248 drop_bs (NonRec _ _) (_ : bs) = bs
249 drop_bs (Rec prs) bs = drop (length prs) bs
251 simpl_bind env bind bs
252 = getDOptsSmpl `thenSmpl` \ dflags ->
253 if dopt Opt_D_dump_inlinings dflags then
254 pprTrace "SimplBind" (ppr (bindersOf bind)) $ simpl_bind1 env bind bs
256 simpl_bind1 env bind bs
258 simpl_bind1 env (NonRec b r) (b':_) = simplRecOrTopPair env TopLevel b b' r
259 simpl_bind1 env (Rec pairs) bs' = simplRecBind env TopLevel pairs bs'
263 %************************************************************************
265 \subsection{simplNonRec}
267 %************************************************************************
269 simplNonRecBind is used for
270 * non-top-level non-recursive lets in expressions
274 * An unsimplified (binder, rhs) pair
275 * The env for the RHS. It may not be the same as the
276 current env because the bind might occur via (\x.E) arg
278 It uses the CPS form because the binding might be strict, in which
279 case we might discard the continuation:
280 let x* = error "foo" in (...x...)
282 It needs to turn unlifted bindings into a @case@. They can arise
283 from, say: (\x -> e) (4# + 3#)
286 simplNonRecBind :: SimplEnv
288 -> InExpr -> SimplEnv -- Arg, with its subst-env
289 -> OutType -- Type of thing computed by the context
290 -> (SimplEnv -> SimplM FloatsWithExpr) -- The body
291 -> SimplM FloatsWithExpr
293 simplNonRecBind env bndr rhs rhs_se cont_ty thing_inside
295 = pprPanic "simplNonRecBind" (ppr bndr <+> ppr rhs)
298 simplNonRecBind env bndr rhs rhs_se cont_ty thing_inside
299 | preInlineUnconditionally env NotTopLevel bndr
300 = tick (PreInlineUnconditionally bndr) `thenSmpl_`
301 thing_inside (extendSubst env bndr (ContEx (getSubstEnv rhs_se) rhs))
304 | isStrictDmd (idNewDemandInfo bndr) || isStrictType (idType bndr) -- A strict let
305 = -- Don't use simplBinder because that doesn't keep
306 -- fragile occurrence info in the substitution
307 simplLetBndr env bndr `thenSmpl` \ (env, bndr1) ->
308 simplStrictArg AnRhs env rhs rhs_se (idType bndr1) cont_ty $ \ env1 rhs1 ->
310 -- Now complete the binding and simplify the body
312 -- simplLetBndr doesn't deal with the IdInfo, so we must
313 -- do so here (c.f. simplLazyBind)
314 bndr2 = bndr1 `setIdInfo` simplIdInfo (getSubst env) (idInfo bndr)
315 env2 = modifyInScope env1 bndr2 bndr2
317 completeNonRecX env2 True {- strict -} bndr bndr2 rhs1 thing_inside
319 | otherwise -- Normal, lazy case
320 = -- Don't use simplBinder because that doesn't keep
321 -- fragile occurrence info in the substitution
322 simplLetBndr env bndr `thenSmpl` \ (env, bndr') ->
323 simplLazyBind env NotTopLevel NonRecursive
324 bndr bndr' rhs rhs_se `thenSmpl` \ (floats, env) ->
325 addFloats env floats thing_inside
328 A specialised variant of simplNonRec used when the RHS is already simplified, notably
329 in knownCon. It uses case-binding where necessary.
332 simplNonRecX :: SimplEnv
333 -> InId -- Old binder
334 -> OutExpr -- Simplified RHS
335 -> (SimplEnv -> SimplM FloatsWithExpr)
336 -> SimplM FloatsWithExpr
338 simplNonRecX env bndr new_rhs thing_inside
339 | needsCaseBinding (idType bndr) new_rhs
340 -- Make this test *before* the preInlineUnconditionally
341 -- Consider case I# (quotInt# x y) of
342 -- I# v -> let w = J# v in ...
343 -- If we gaily inline (quotInt# x y) for v, we end up building an
345 -- let w = J# (quotInt# x y) in ...
346 -- because quotInt# can fail.
347 = simplBinder env bndr `thenSmpl` \ (env, bndr') ->
348 thing_inside env `thenSmpl` \ (floats, body) ->
349 returnSmpl (emptyFloats env, Case new_rhs bndr' [(DEFAULT, [], wrapFloats floats body)])
351 | preInlineUnconditionally env NotTopLevel bndr
352 -- This happens; for example, the case_bndr during case of
353 -- known constructor: case (a,b) of x { (p,q) -> ... }
354 -- Here x isn't mentioned in the RHS, so we don't want to
355 -- create the (dead) let-binding let x = (a,b) in ...
357 -- Similarly, single occurrences can be inlined vigourously
358 -- e.g. case (f x, g y) of (a,b) -> ....
359 -- If a,b occur once we can avoid constructing the let binding for them.
360 = thing_inside (extendSubst env bndr (ContEx emptySubstEnv new_rhs))
363 = simplBinder env bndr `thenSmpl` \ (env, bndr') ->
364 completeNonRecX env False {- Non-strict; pessimistic -}
365 bndr bndr' new_rhs thing_inside
367 completeNonRecX env is_strict old_bndr new_bndr new_rhs thing_inside
368 = mkAtomicArgs is_strict
369 True {- OK to float unlifted -}
370 new_rhs `thenSmpl` \ (aux_binds, rhs2) ->
372 -- Make the arguments atomic if necessary,
373 -- adding suitable bindings
374 addAtomicBindsE env (fromOL aux_binds) $ \ env ->
375 completeLazyBind env NotTopLevel
376 old_bndr new_bndr rhs2 `thenSmpl` \ (floats, env) ->
377 addFloats env floats thing_inside
381 %************************************************************************
383 \subsection{Lazy bindings}
385 %************************************************************************
387 simplRecBind is used for
388 * recursive bindings only
391 simplRecBind :: SimplEnv -> TopLevelFlag
392 -> [(InId, InExpr)] -> [OutId]
393 -> SimplM (FloatsWith SimplEnv)
394 simplRecBind env top_lvl pairs bndrs'
395 = go env pairs bndrs' `thenSmpl` \ (floats, env) ->
396 returnSmpl (flattenFloats floats, env)
398 go env [] _ = returnSmpl (emptyFloats env, env)
400 go env ((bndr, rhs) : pairs) (bndr' : bndrs')
401 = simplRecOrTopPair env top_lvl bndr bndr' rhs `thenSmpl` \ (floats, env) ->
402 addFloats env floats (\env -> go env pairs bndrs')
406 simplRecOrTopPair is used for
407 * recursive bindings (whether top level or not)
408 * top-level non-recursive bindings
410 It assumes the binder has already been simplified, but not its IdInfo.
413 simplRecOrTopPair :: SimplEnv
415 -> InId -> OutId -- Binder, both pre-and post simpl
416 -> InExpr -- The RHS and its environment
417 -> SimplM (FloatsWith SimplEnv)
419 simplRecOrTopPair env top_lvl bndr bndr' rhs
420 | preInlineUnconditionally env top_lvl bndr -- Check for unconditional inline
421 = tick (PreInlineUnconditionally bndr) `thenSmpl_`
422 returnSmpl (emptyFloats env, extendSubst env bndr (ContEx (getSubstEnv env) rhs))
425 = simplLazyBind env top_lvl Recursive bndr bndr' rhs env
426 -- May not actually be recursive, but it doesn't matter
430 simplLazyBind is used for
431 * recursive bindings (whether top level or not)
432 * top-level non-recursive bindings
433 * non-top-level *lazy* non-recursive bindings
435 [Thus it deals with the lazy cases from simplNonRecBind, and all cases
436 from SimplRecOrTopBind]
439 1. It assumes that the binder is *already* simplified,
440 and is in scope, but not its IdInfo
442 2. It assumes that the binder type is lifted.
444 3. It does not check for pre-inline-unconditionallly;
445 that should have been done already.
448 simplLazyBind :: SimplEnv
449 -> TopLevelFlag -> RecFlag
450 -> InId -> OutId -- Binder, both pre-and post simpl
451 -> InExpr -> SimplEnv -- The RHS and its environment
452 -> SimplM (FloatsWith SimplEnv)
454 simplLazyBind env top_lvl is_rec bndr bndr1 rhs rhs_se
455 = let -- Transfer the IdInfo of the original binder to the new binder
456 -- This is crucial: we must preserve
460 -- etc. To do this we must apply the current substitution,
461 -- which incorporates earlier substitutions in this very letrec group.
463 -- NB 1. We do this *before* processing the RHS of the binder, so that
464 -- its substituted rules are visible in its own RHS.
465 -- This is important. Manuel found cases where he really, really
466 -- wanted a RULE for a recursive function to apply in that function's
467 -- own right-hand side.
469 -- NB 2: We do not transfer the arity (see Subst.substIdInfo)
470 -- The arity of an Id should not be visible
471 -- in its own RHS, else we eta-reduce
475 -- which isn't sound. And it makes the arity in f's IdInfo greater than
476 -- the manifest arity, which isn't good.
477 -- The arity will get added later.
479 -- NB 3: It's important that we *do* transer the loop-breaker OccInfo,
480 -- because that's what stops the Id getting inlined infinitely, in the body
483 -- NB 4: does no harm for non-recursive bindings
485 bndr2 = bndr1 `setIdInfo` simplIdInfo (getSubst env) (idInfo bndr)
486 env1 = modifyInScope env bndr2 bndr2
487 rhs_env = setInScope rhs_se env1
488 is_top_level = isTopLevel top_lvl
489 ok_float_unlifted = not is_top_level && isNonRec is_rec
490 rhs_cont = mkStop (idType bndr1) AnRhs
492 -- Simplify the RHS; note the mkStop, which tells
493 -- the simplifier that this is the RHS of a let.
494 simplExprF rhs_env rhs rhs_cont `thenSmpl` \ (floats, rhs1) ->
496 -- If any of the floats can't be floated, give up now
497 -- (The allLifted predicate says True for empty floats.)
498 if (not ok_float_unlifted && not (allLifted floats)) then
499 completeLazyBind env1 top_lvl bndr bndr2
500 (wrapFloats floats rhs1)
503 -- ANF-ise a constructor or PAP rhs
504 mkAtomicArgs False {- Not strict -}
505 ok_float_unlifted rhs1 `thenSmpl` \ (aux_binds, rhs2) ->
507 -- If the result is a PAP, float the floats out, else wrap them
508 -- By this time it's already been ANF-ised (if necessary)
509 if isEmptyFloats floats && isNilOL aux_binds then -- Shortcut a common case
510 completeLazyBind env1 top_lvl bndr bndr2 rhs2
512 -- We use exprIsTrivial here because we want to reveal lone variables.
513 -- E.g. let { x = letrec { y = E } in y } in ...
514 -- Here we definitely want to float the y=E defn.
515 -- exprIsValue definitely isn't right for that.
517 -- BUT we can't use "exprIsCheap", because that causes a strictness bug.
518 -- x = let y* = E in case (scc y) of { T -> F; F -> T}
519 -- The case expression is 'cheap', but it's wrong to transform to
520 -- y* = E; x = case (scc y) of {...}
521 -- Either we must be careful not to float demanded non-values, or
522 -- we must use exprIsValue for the test, which ensures that the
523 -- thing is non-strict. I think. The WARN below tests for this.
524 else if is_top_level || exprIsTrivial rhs2 || exprIsValue rhs2 then
526 -- There's a subtlety here. There may be a binding (x* = e) in the
527 -- floats, where the '*' means 'will be demanded'. So is it safe
528 -- to float it out? Answer no, but it won't matter because
529 -- we only float if (a) arg' is a WHNF, or (b) it's going to top level
530 -- and so there can't be any 'will be demanded' bindings in the floats.
532 WARN( not is_top_level && any demanded_float (floatBinds floats),
533 ppr (filter demanded_float (floatBinds floats)) )
535 tick LetFloatFromLet `thenSmpl_` (
536 addFloats env1 floats $ \ env2 ->
537 addAtomicBinds env2 (fromOL aux_binds) $ \ env3 ->
538 completeLazyBind env3 top_lvl bndr bndr2 rhs2)
541 completeLazyBind env1 top_lvl bndr bndr2 (wrapFloats floats rhs1)
544 demanded_float (NonRec b r) = isStrictDmd (idNewDemandInfo b) && not (isUnLiftedType (idType b))
545 -- Unlifted-type (cheap-eagerness) lets may well have a demanded flag on them
546 demanded_float (Rec _) = False
551 %************************************************************************
553 \subsection{Completing a lazy binding}
555 %************************************************************************
558 * deals only with Ids, not TyVars
559 * takes an already-simplified binder and RHS
560 * is used for both recursive and non-recursive bindings
561 * is used for both top-level and non-top-level bindings
563 It does the following:
564 - tries discarding a dead binding
565 - tries PostInlineUnconditionally
566 - add unfolding [this is the only place we add an unfolding]
569 It does *not* attempt to do let-to-case. Why? Because it is used for
570 - top-level bindings (when let-to-case is impossible)
571 - many situations where the "rhs" is known to be a WHNF
572 (so let-to-case is inappropriate).
575 completeLazyBind :: SimplEnv
576 -> TopLevelFlag -- Flag stuck into unfolding
577 -> InId -- Old binder
578 -> OutId -- New binder
579 -> OutExpr -- Simplified RHS
580 -> SimplM (FloatsWith SimplEnv)
581 -- We return a new SimplEnv, because completeLazyBind may choose to do its work
582 -- by extending the substitution (e.g. let x = y in ...)
583 -- The new binding (if any) is returned as part of the floats.
584 -- NB: the returned SimplEnv has the right SubstEnv, but you should
585 -- (as usual) use the in-scope-env from the floats
587 completeLazyBind env top_lvl old_bndr new_bndr new_rhs
588 | postInlineUnconditionally env new_bndr occ_info new_rhs
589 = -- Drop the binding
590 tick (PostInlineUnconditionally old_bndr) `thenSmpl_`
591 returnSmpl (emptyFloats env, extendSubst env old_bndr (DoneEx new_rhs))
592 -- Use the substitution to make quite, quite sure that the substitution
593 -- will happen, since we are going to discard the binding
598 new_bndr_info = idInfo new_bndr `setArityInfo` exprArity new_rhs
600 -- Add the unfolding *only* for non-loop-breakers
601 -- Making loop breakers not have an unfolding at all
602 -- means that we can avoid tests in exprIsConApp, for example.
603 -- This is important: if exprIsConApp says 'yes' for a recursive
604 -- thing, then we can get into an infinite loop
605 info_w_unf | loop_breaker = new_bndr_info
606 | otherwise = new_bndr_info `setUnfoldingInfo` unfolding
607 unfolding = mkUnfolding (isTopLevel top_lvl) new_rhs
609 final_id = new_bndr `setIdInfo` info_w_unf
611 -- These seqs forces the Id, and hence its IdInfo,
612 -- and hence any inner substitutions
614 returnSmpl (unitFloat env final_id new_rhs, env)
617 loop_breaker = isLoopBreaker occ_info
618 old_info = idInfo old_bndr
619 occ_info = occInfo old_info
624 %************************************************************************
626 \subsection[Simplify-simplExpr]{The main function: simplExpr}
628 %************************************************************************
630 The reason for this OutExprStuff stuff is that we want to float *after*
631 simplifying a RHS, not before. If we do so naively we get quadratic
632 behaviour as things float out.
634 To see why it's important to do it after, consider this (real) example:
648 a -- Can't inline a this round, cos it appears twice
652 Each of the ==> steps is a round of simplification. We'd save a
653 whole round if we float first. This can cascade. Consider
658 let f = let d1 = ..d.. in \y -> e
662 in \x -> ...(\y ->e)...
664 Only in this second round can the \y be applied, and it
665 might do the same again.
669 simplExpr :: SimplEnv -> CoreExpr -> SimplM CoreExpr
670 simplExpr env expr = simplExprC env expr (mkStop expr_ty' AnArg)
672 expr_ty' = substTy (getSubst env) (exprType expr)
673 -- The type in the Stop continuation, expr_ty', is usually not used
674 -- It's only needed when discarding continuations after finding
675 -- a function that returns bottom.
676 -- Hence the lazy substitution
679 simplExprC :: SimplEnv -> CoreExpr -> SimplCont -> SimplM CoreExpr
680 -- Simplify an expression, given a continuation
681 simplExprC env expr cont
682 = simplExprF env expr cont `thenSmpl` \ (floats, expr) ->
683 returnSmpl (wrapFloats floats expr)
685 simplExprF :: SimplEnv -> InExpr -> SimplCont -> SimplM FloatsWithExpr
686 -- Simplify an expression, returning floated binds
688 simplExprF env (Var v) cont = simplVar env v cont
689 simplExprF env (Lit lit) cont = rebuild env (Lit lit) cont
690 simplExprF env expr@(Lam _ _) cont = simplLam env expr cont
691 simplExprF env (Note note expr) cont = simplNote env note expr cont
692 simplExprF env (App fun arg) cont = simplExprF env fun (ApplyTo NoDup arg env cont)
694 simplExprF env (Type ty) cont
695 = ASSERT( contIsRhsOrArg cont )
696 simplType env ty `thenSmpl` \ ty' ->
697 rebuild env (Type ty') cont
699 simplExprF env (Case scrut bndr alts) cont
700 | not (switchIsOn (getSwitchChecker env) NoCaseOfCase)
701 = -- Simplify the scrutinee with a Select continuation
702 simplExprF env scrut (Select NoDup bndr alts env cont)
705 = -- If case-of-case is off, simply simplify the case expression
706 -- in a vanilla Stop context, and rebuild the result around it
707 simplExprC env scrut case_cont `thenSmpl` \ case_expr' ->
708 rebuild env case_expr' cont
710 case_cont = Select NoDup bndr alts env (mkBoringStop (contResultType cont))
712 simplExprF env (Let (Rec pairs) body) cont
713 = simplRecBndrs env (map fst pairs) `thenSmpl` \ (env, bndrs') ->
714 -- NB: bndrs' don't have unfoldings or rules
715 -- We add them as we go down
717 simplRecBind env NotTopLevel pairs bndrs' `thenSmpl` \ (floats, env) ->
718 addFloats env floats $ \ env ->
719 simplExprF env body cont
721 -- A non-recursive let is dealt with by simplNonRecBind
722 simplExprF env (Let (NonRec bndr rhs) body) cont
723 = simplNonRecBind env bndr rhs env (contResultType cont) $ \ env ->
724 simplExprF env body cont
727 ---------------------------------
728 simplType :: SimplEnv -> InType -> SimplM OutType
729 -- Kept monadic just so we can do the seqType
731 = seqType new_ty `seq` returnSmpl new_ty
733 new_ty = substTy (getSubst env) ty
737 %************************************************************************
741 %************************************************************************
744 simplLam env fun cont
747 zap_it = mkLamBndrZapper fun (countArgs cont)
748 cont_ty = contResultType cont
750 -- Type-beta reduction
751 go env (Lam bndr body) (ApplyTo _ (Type ty_arg) arg_se body_cont)
752 = ASSERT( isTyVar bndr )
753 tick (BetaReduction bndr) `thenSmpl_`
754 simplType (setInScope arg_se env) ty_arg `thenSmpl` \ ty_arg' ->
755 go (extendSubst env bndr (DoneTy ty_arg')) body body_cont
757 -- Ordinary beta reduction
758 go env (Lam bndr body) cont@(ApplyTo _ arg arg_se body_cont)
759 = tick (BetaReduction bndr) `thenSmpl_`
760 simplNonRecBind env (zap_it bndr) arg arg_se cont_ty $ \ env ->
761 go env body body_cont
763 -- Not enough args, so there are real lambdas left to put in the result
764 go env lam@(Lam _ _) cont
765 = simplLamBndrs env bndrs `thenSmpl` \ (env, bndrs') ->
766 simplExpr env body `thenSmpl` \ body' ->
767 mkLam env bndrs' body' cont `thenSmpl` \ (floats, new_lam) ->
768 addFloats env floats $ \ env ->
769 rebuild env new_lam cont
771 (bndrs,body) = collectBinders lam
773 -- Exactly enough args
774 go env expr cont = simplExprF env expr cont
776 mkLamBndrZapper :: CoreExpr -- Function
777 -> Int -- Number of args supplied, *including* type args
778 -> Id -> Id -- Use this to zap the binders
779 mkLamBndrZapper fun n_args
780 | n_args >= n_params fun = \b -> b -- Enough args
781 | otherwise = \b -> zapLamIdInfo b
783 -- NB: we count all the args incl type args
784 -- so we must count all the binders (incl type lambdas)
785 n_params (Note _ e) = n_params e
786 n_params (Lam b e) = 1 + n_params e
787 n_params other = 0::Int
791 %************************************************************************
795 %************************************************************************
798 simplNote env (Coerce to from) body cont
800 in_scope = getInScope env
802 addCoerce s1 k1 (CoerceIt t1 cont)
803 -- coerce T1 S1 (coerce S1 K1 e)
806 -- coerce T1 K1 e, otherwise
808 -- For example, in the initial form of a worker
809 -- we may find (coerce T (coerce S (\x.e))) y
810 -- and we'd like it to simplify to e[y/x] in one round
812 | t1 `eqType` k1 = cont -- The coerces cancel out
813 | otherwise = CoerceIt t1 cont -- They don't cancel, but
814 -- the inner one is redundant
816 addCoerce t1t2 s1s2 (ApplyTo dup arg arg_se cont)
817 | not (isTypeArg arg), -- This whole case only works for value args
818 -- Could upgrade to have equiv thing for type apps too
819 Just (s1, s2) <- splitFunTy_maybe s1s2
820 -- (coerce (T1->T2) (S1->S2) F) E
822 -- coerce T2 S2 (F (coerce S1 T1 E))
824 -- t1t2 must be a function type, T1->T2, because it's applied to something
825 -- but s1s2 might conceivably not be
827 -- When we build the ApplyTo we can't mix the out-types
828 -- with the InExpr in the argument, so we simply substitute
829 -- to make it all consistent. It's a bit messy.
830 -- But it isn't a common case.
832 (t1,t2) = splitFunTy t1t2
833 new_arg = mkCoerce2 s1 t1 (substExpr (mkSubst in_scope (getSubstEnv arg_se)) arg)
835 ApplyTo dup new_arg (zapSubstEnv env) (addCoerce t2 s2 cont)
837 addCoerce to' _ cont = CoerceIt to' cont
839 simplType env to `thenSmpl` \ to' ->
840 simplType env from `thenSmpl` \ from' ->
841 simplExprF env body (addCoerce to' from' cont)
844 -- Hack: we only distinguish subsumed cost centre stacks for the purposes of
845 -- inlining. All other CCCSs are mapped to currentCCS.
846 simplNote env (SCC cc) e cont
847 = simplExpr (setEnclosingCC env currentCCS) e `thenSmpl` \ e' ->
848 rebuild env (mkSCC cc e') cont
850 simplNote env InlineCall e cont
851 = simplExprF env e (InlinePlease cont)
853 -- See notes with SimplMonad.inlineMode
854 simplNote env InlineMe e cont
855 | contIsRhsOrArg cont -- Totally boring continuation; see notes above
856 = -- Don't inline inside an INLINE expression
857 simplExpr (setMode inlineMode env ) e `thenSmpl` \ e' ->
858 rebuild env (mkInlineMe e') cont
860 | otherwise -- Dissolve the InlineMe note if there's
861 -- an interesting context of any kind to combine with
862 -- (even a type application -- anything except Stop)
863 = simplExprF env e cont
865 simplNote env (CoreNote s) e cont
866 = simplExpr env e `thenSmpl` \ e' ->
867 rebuild env (Note (CoreNote s) e') cont
871 %************************************************************************
873 \subsection{Dealing with calls}
875 %************************************************************************
878 simplVar env var cont
879 = case lookupIdSubst (getSubst env) var of
880 DoneEx e -> simplExprF (zapSubstEnv env) e cont
881 ContEx se e -> simplExprF (setSubstEnv env se) e cont
882 DoneId var1 occ -> WARN( not (isInScope var1 (getSubst env)) && mustHaveLocalBinding var1,
883 text "simplVar:" <+> ppr var )
884 completeCall (zapSubstEnv env) var1 occ cont
885 -- The template is already simplified, so don't re-substitute.
886 -- This is VITAL. Consider
888 -- let y = \z -> ...x... in
890 -- We'll clone the inner \x, adding x->x' in the id_subst
891 -- Then when we inline y, we must *not* replace x by x' in
892 -- the inlined copy!!
894 ---------------------------------------------------------
895 -- Dealing with a call site
897 completeCall env var occ_info cont
898 = -- Simplify the arguments
899 getDOptsSmpl `thenSmpl` \ dflags ->
901 chkr = getSwitchChecker env
902 (args, call_cont, inline_call) = getContArgs chkr var cont
905 simplifyArgs env fn_ty args (contResultType call_cont) $ \ env args ->
907 -- Next, look for rules or specialisations that match
909 -- It's important to simplify the args first, because the rule-matcher
910 -- doesn't do substitution as it goes. We don't want to use subst_args
911 -- (defined in the 'where') because that throws away useful occurrence info,
912 -- and perhaps-very-important specialisations.
914 -- Some functions have specialisations *and* are strict; in this case,
915 -- we don't want to inline the wrapper of the non-specialised thing; better
916 -- to call the specialised thing instead.
917 -- We used to use the black-listing mechanism to ensure that inlining of
918 -- the wrapper didn't occur for things that have specialisations till a
919 -- later phase, so but now we just try RULES first
921 -- You might think that we shouldn't apply rules for a loop breaker:
922 -- doing so might give rise to an infinite loop, because a RULE is
923 -- rather like an extra equation for the function:
924 -- RULE: f (g x) y = x+y
927 -- But it's too drastic to disable rules for loop breakers.
928 -- Even the foldr/build rule would be disabled, because foldr
929 -- is recursive, and hence a loop breaker:
930 -- foldr k z (build g) = g k z
931 -- So it's up to the programmer: rules can cause divergence
934 in_scope = getInScope env
935 maybe_rule = case activeRule env of
936 Nothing -> Nothing -- No rules apply
937 Just act_fn -> lookupRule act_fn in_scope var args
940 Just (rule_name, rule_rhs) ->
941 tick (RuleFired rule_name) `thenSmpl_`
942 (if dopt Opt_D_dump_inlinings dflags then
943 pprTrace "Rule fired" (vcat [
944 text "Rule:" <+> ftext rule_name,
945 text "Before:" <+> ppr var <+> sep (map pprParendExpr args),
946 text "After: " <+> pprCoreExpr rule_rhs,
947 text "Cont: " <+> ppr call_cont])
950 simplExprF env rule_rhs call_cont ;
952 Nothing -> -- No rules
954 -- Next, look for an inlining
956 arg_infos = [ interestingArg arg | arg <- args, isValArg arg]
958 interesting_cont = interestingCallContext (notNull args)
962 active_inline = activeInline env var occ_info
963 maybe_inline = callSiteInline dflags active_inline inline_call occ_info
964 var arg_infos interesting_cont
966 case maybe_inline of {
967 Just unfolding -- There is an inlining!
968 -> tick (UnfoldingDone var) `thenSmpl_`
969 makeThatCall env var unfolding args call_cont
972 Nothing -> -- No inlining!
975 rebuild env (mkApps (Var var) args) call_cont
978 makeThatCall :: SimplEnv
980 -> InExpr -- Inlined function rhs
981 -> [OutExpr] -- Arguments, already simplified
982 -> SimplCont -- After the call
983 -> SimplM FloatsWithExpr
984 -- Similar to simplLam, but this time
985 -- the arguments are already simplified
986 makeThatCall orig_env var fun@(Lam _ _) args cont
987 = go orig_env fun args
989 zap_it = mkLamBndrZapper fun (length args)
991 -- Type-beta reduction
992 go env (Lam bndr body) (Type ty_arg : args)
993 = ASSERT( isTyVar bndr )
994 tick (BetaReduction bndr) `thenSmpl_`
995 go (extendSubst env bndr (DoneTy ty_arg)) body args
997 -- Ordinary beta reduction
998 go env (Lam bndr body) (arg : args)
999 = tick (BetaReduction bndr) `thenSmpl_`
1000 simplNonRecX env (zap_it bndr) arg $ \ env ->
1003 -- Not enough args, so there are real lambdas left to put in the result
1005 = simplExprF env fun (pushContArgs orig_env args cont)
1006 -- NB: orig_env; the correct environment to capture with
1007 -- the arguments.... env has been augmented with substitutions
1008 -- from the beta reductions.
1010 makeThatCall env var fun args cont
1011 = simplExprF env fun (pushContArgs env args cont)
1015 %************************************************************************
1017 \subsection{Arguments}
1019 %************************************************************************
1022 ---------------------------------------------------------
1023 -- Simplifying the arguments of a call
1025 simplifyArgs :: SimplEnv
1026 -> OutType -- Type of the function
1027 -> [(InExpr, SimplEnv, Bool)] -- Details of the arguments
1028 -> OutType -- Type of the continuation
1029 -> (SimplEnv -> [OutExpr] -> SimplM FloatsWithExpr)
1030 -> SimplM FloatsWithExpr
1032 -- [CPS-like because of strict arguments]
1034 -- Simplify the arguments to a call.
1035 -- This part of the simplifier may break the no-shadowing invariant
1037 -- f (...(\a -> e)...) (case y of (a,b) -> e')
1038 -- where f is strict in its second arg
1039 -- If we simplify the innermost one first we get (...(\a -> e)...)
1040 -- Simplifying the second arg makes us float the case out, so we end up with
1041 -- case y of (a,b) -> f (...(\a -> e)...) e'
1042 -- So the output does not have the no-shadowing invariant. However, there is
1043 -- no danger of getting name-capture, because when the first arg was simplified
1044 -- we used an in-scope set that at least mentioned all the variables free in its
1045 -- static environment, and that is enough.
1047 -- We can't just do innermost first, or we'd end up with a dual problem:
1048 -- case x of (a,b) -> f e (...(\a -> e')...)
1050 -- I spent hours trying to recover the no-shadowing invariant, but I just could
1051 -- not think of an elegant way to do it. The simplifier is already knee-deep in
1052 -- continuations. We have to keep the right in-scope set around; AND we have
1053 -- to get the effect that finding (error "foo") in a strict arg position will
1054 -- discard the entire application and replace it with (error "foo"). Getting
1055 -- all this at once is TOO HARD!
1057 simplifyArgs env fn_ty args cont_ty thing_inside
1058 = go env fn_ty args thing_inside
1060 go env fn_ty [] thing_inside = thing_inside env []
1061 go env fn_ty (arg:args) thing_inside = simplifyArg env fn_ty arg cont_ty $ \ env arg' ->
1062 go env (applyTypeToArg fn_ty arg') args $ \ env args' ->
1063 thing_inside env (arg':args')
1065 simplifyArg env fn_ty (Type ty_arg, se, _) cont_ty thing_inside
1066 = simplType (setInScope se env) ty_arg `thenSmpl` \ new_ty_arg ->
1067 thing_inside env (Type new_ty_arg)
1069 simplifyArg env fn_ty (val_arg, arg_se, is_strict) cont_ty thing_inside
1071 = simplStrictArg AnArg env val_arg arg_se arg_ty cont_ty thing_inside
1073 | otherwise -- Lazy argument
1074 -- DO NOT float anything outside, hence simplExprC
1075 -- There is no benefit (unlike in a let-binding), and we'd
1076 -- have to be very careful about bogus strictness through
1077 -- floating a demanded let.
1078 = simplExprC (setInScope arg_se env) val_arg
1079 (mkStop arg_ty AnArg) `thenSmpl` \ arg1 ->
1080 thing_inside env arg1
1082 arg_ty = funArgTy fn_ty
1085 simplStrictArg :: LetRhsFlag
1086 -> SimplEnv -- The env of the call
1087 -> InExpr -> SimplEnv -- The arg plus its env
1088 -> OutType -- arg_ty: type of the argument
1089 -> OutType -- cont_ty: Type of thing computed by the context
1090 -> (SimplEnv -> OutExpr -> SimplM FloatsWithExpr)
1091 -- Takes an expression of type rhs_ty,
1092 -- returns an expression of type cont_ty
1093 -- The env passed to this continuation is the
1094 -- env of the call, plus any new in-scope variables
1095 -> SimplM FloatsWithExpr -- An expression of type cont_ty
1097 simplStrictArg is_rhs call_env arg arg_env arg_ty cont_ty thing_inside
1098 = simplExprF (setInScope arg_env call_env) arg
1099 (ArgOf is_rhs arg_ty cont_ty (\ new_env -> thing_inside (setInScope call_env new_env)))
1100 -- Notice the way we use arg_env (augmented with in-scope vars from call_env)
1101 -- to simplify the argument
1102 -- and call-env (augmented with in-scope vars from the arg) to pass to the continuation
1106 %************************************************************************
1108 \subsection{mkAtomicArgs}
1110 %************************************************************************
1112 mkAtomicArgs takes a putative RHS, checks whether it's a PAP or
1113 constructor application and, if so, converts it to ANF, so that the
1114 resulting thing can be inlined more easily. Thus
1121 There are three sorts of binding context, specified by the two
1127 N N Top-level or recursive Only bind args of lifted type
1129 N Y Non-top-level and non-recursive, Bind args of lifted type, or
1130 but lazy unlifted-and-ok-for-speculation
1132 Y Y Non-top-level, non-recursive, Bind all args
1133 and strict (demanded)
1140 there is no point in transforming to
1142 x = case (y div# z) of r -> MkC r
1144 because the (y div# z) can't float out of the let. But if it was
1145 a *strict* let, then it would be a good thing to do. Hence the
1146 context information.
1149 mkAtomicArgs :: Bool -- A strict binding
1150 -> Bool -- OK to float unlifted args
1152 -> SimplM (OrdList (OutId,OutExpr), -- The floats (unusually) may include
1153 OutExpr) -- things that need case-binding,
1154 -- if the strict-binding flag is on
1156 mkAtomicArgs is_strict ok_float_unlifted rhs
1157 | (Var fun, args) <- collectArgs rhs, -- It's an application
1158 isDataConWorkId fun || valArgCount args < idArity fun -- And it's a constructor or PAP
1159 = go fun nilOL [] args -- Have a go
1161 | otherwise = bale_out -- Give up
1164 bale_out = returnSmpl (nilOL, rhs)
1166 go fun binds rev_args []
1167 = returnSmpl (binds, mkApps (Var fun) (reverse rev_args))
1169 go fun binds rev_args (arg : args)
1170 | exprIsTrivial arg -- Easy case
1171 = go fun binds (arg:rev_args) args
1173 | not can_float_arg -- Can't make this arg atomic
1174 = bale_out -- ... so give up
1176 | otherwise -- Don't forget to do it recursively
1177 -- E.g. x = a:b:c:[]
1178 = mkAtomicArgs is_strict ok_float_unlifted arg `thenSmpl` \ (arg_binds, arg') ->
1179 newId FSLIT("a") arg_ty `thenSmpl` \ arg_id ->
1180 go fun ((arg_binds `snocOL` (arg_id,arg')) `appOL` binds)
1181 (Var arg_id : rev_args) args
1183 arg_ty = exprType arg
1184 can_float_arg = is_strict
1185 || not (isUnLiftedType arg_ty)
1186 || (ok_float_unlifted && exprOkForSpeculation arg)
1189 addAtomicBinds :: SimplEnv -> [(OutId,OutExpr)]
1190 -> (SimplEnv -> SimplM (FloatsWith a))
1191 -> SimplM (FloatsWith a)
1192 addAtomicBinds env [] thing_inside = thing_inside env
1193 addAtomicBinds env ((v,r):bs) thing_inside = addAuxiliaryBind env (NonRec v r) $ \ env ->
1194 addAtomicBinds env bs thing_inside
1196 addAtomicBindsE :: SimplEnv -> [(OutId,OutExpr)]
1197 -> (SimplEnv -> SimplM FloatsWithExpr)
1198 -> SimplM FloatsWithExpr
1199 -- Same again, but this time we're in an expression context,
1200 -- and may need to do some case bindings
1202 addAtomicBindsE env [] thing_inside
1204 addAtomicBindsE env ((v,r):bs) thing_inside
1205 | needsCaseBinding (idType v) r
1206 = addAtomicBindsE (addNewInScopeIds env [v]) bs thing_inside `thenSmpl` \ (floats, expr) ->
1207 WARN( exprIsTrivial expr, ppr v <+> pprCoreExpr expr )
1208 returnSmpl (emptyFloats env, Case r v [(DEFAULT,[], wrapFloats floats expr)])
1211 = addAuxiliaryBind env (NonRec v r) $ \ env ->
1212 addAtomicBindsE env bs thing_inside
1216 %************************************************************************
1218 \subsection{The main rebuilder}
1220 %************************************************************************
1223 rebuild :: SimplEnv -> OutExpr -> SimplCont -> SimplM FloatsWithExpr
1225 rebuild env expr (Stop _ _ _) = rebuildDone env expr
1226 rebuild env expr (ArgOf _ _ _ cont_fn) = cont_fn env expr
1227 rebuild env expr (CoerceIt to_ty cont) = rebuild env (mkCoerce to_ty expr) cont
1228 rebuild env expr (InlinePlease cont) = rebuild env (Note InlineCall expr) cont
1229 rebuild env expr (Select _ bndr alts se cont) = rebuildCase (setInScope se env) expr bndr alts cont
1230 rebuild env expr (ApplyTo _ arg se cont) = rebuildApp (setInScope se env) expr arg cont
1232 rebuildApp env fun arg cont
1233 = simplExpr env arg `thenSmpl` \ arg' ->
1234 rebuild env (App fun arg') cont
1236 rebuildDone env expr = returnSmpl (emptyFloats env, expr)
1240 %************************************************************************
1242 \subsection{Functions dealing with a case}
1244 %************************************************************************
1246 Blob of helper functions for the "case-of-something-else" situation.
1249 ---------------------------------------------------------
1250 -- Eliminate the case if possible
1252 rebuildCase :: SimplEnv
1253 -> OutExpr -- Scrutinee
1254 -> InId -- Case binder
1255 -> [InAlt] -- Alternatives
1257 -> SimplM FloatsWithExpr
1259 rebuildCase env scrut case_bndr alts cont
1260 | Just (con,args) <- exprIsConApp_maybe scrut
1261 -- Works when the scrutinee is a variable with a known unfolding
1262 -- as well as when it's an explicit constructor application
1263 = knownCon env (DataAlt con) args case_bndr alts cont
1265 | Lit lit <- scrut -- No need for same treatment as constructors
1266 -- because literals are inlined more vigorously
1267 = knownCon env (LitAlt lit) [] case_bndr alts cont
1270 = prepareAlts scrut case_bndr alts `thenSmpl` \ (better_alts, handled_cons) ->
1272 -- Deal with the case binder, and prepare the continuation;
1273 -- The new subst_env is in place
1274 prepareCaseCont env better_alts cont `thenSmpl` \ (floats, (dup_cont, nondup_cont)) ->
1275 addFloats env floats $ \ env ->
1277 -- Deal with variable scrutinee
1278 simplCaseBinder env scrut case_bndr `thenSmpl` \ (alt_env, case_bndr', zap_occ_info) ->
1280 -- Deal with the case alternatives
1281 simplAlts alt_env zap_occ_info handled_cons
1282 case_bndr' better_alts dup_cont `thenSmpl` \ alts' ->
1284 -- Put the case back together
1285 mkCase scrut case_bndr' alts' `thenSmpl` \ case_expr ->
1287 -- Notice that rebuildDone returns the in-scope set from env, not alt_env
1288 -- The case binder *not* scope over the whole returned case-expression
1289 rebuild env case_expr nondup_cont
1292 simplCaseBinder checks whether the scrutinee is a variable, v. If so,
1293 try to eliminate uses of v in the RHSs in favour of case_bndr; that
1294 way, there's a chance that v will now only be used once, and hence
1299 There is a time we *don't* want to do that, namely when
1300 -fno-case-of-case is on. This happens in the first simplifier pass,
1301 and enhances full laziness. Here's the bad case:
1302 f = \ y -> ...(case x of I# v -> ...(case x of ...) ... )
1303 If we eliminate the inner case, we trap it inside the I# v -> arm,
1304 which might prevent some full laziness happening. I've seen this
1305 in action in spectral/cichelli/Prog.hs:
1306 [(m,n) | m <- [1..max], n <- [1..max]]
1307 Hence the check for NoCaseOfCase.
1311 There is another situation when we don't want to do it. If we have
1313 case x of w1 { DEFAULT -> case x of w2 { A -> e1; B -> e2 }
1314 ...other cases .... }
1316 We'll perform the binder-swap for the outer case, giving
1318 case x of w1 { DEFAULT -> case w1 of w2 { A -> e1; B -> e2 }
1319 ...other cases .... }
1321 But there is no point in doing it for the inner case, because w1 can't
1322 be inlined anyway. Furthermore, doing the case-swapping involves
1323 zapping w2's occurrence info (see paragraphs that follow), and that
1324 forces us to bind w2 when doing case merging. So we get
1326 case x of w1 { A -> let w2 = w1 in e1
1327 B -> let w2 = w1 in e2
1328 ...other cases .... }
1330 This is plain silly in the common case where w2 is dead.
1332 Even so, I can't see a good way to implement this idea. I tried
1333 not doing the binder-swap if the scrutinee was already evaluated
1334 but that failed big-time:
1338 case v of w { MkT x ->
1339 case x of x1 { I# y1 ->
1340 case x of x2 { I# y2 -> ...
1342 Notice that because MkT is strict, x is marked "evaluated". But to
1343 eliminate the last case, we must either make sure that x (as well as
1344 x1) has unfolding MkT y1. THe straightforward thing to do is to do
1345 the binder-swap. So this whole note is a no-op.
1349 If we replace the scrutinee, v, by tbe case binder, then we have to nuke
1350 any occurrence info (eg IAmDead) in the case binder, because the
1351 case-binder now effectively occurs whenever v does. AND we have to do
1352 the same for the pattern-bound variables! Example:
1354 (case x of { (a,b) -> a }) (case x of { (p,q) -> q })
1356 Here, b and p are dead. But when we move the argment inside the first
1357 case RHS, and eliminate the second case, we get
1359 case x or { (a,b) -> a b }
1361 Urk! b is alive! Reason: the scrutinee was a variable, and case elimination
1362 happened. Hence the zap_occ_info function returned by simplCaseBinder
1365 simplCaseBinder env (Var v) case_bndr
1366 | not (switchIsOn (getSwitchChecker env) NoCaseOfCase)
1368 -- Failed try [see Note 2 above]
1369 -- not (isEvaldUnfolding (idUnfolding v))
1371 = simplBinder env (zap case_bndr) `thenSmpl` \ (env, case_bndr') ->
1372 returnSmpl (modifyInScope env v case_bndr', case_bndr', zap)
1373 -- We could extend the substitution instead, but it would be
1374 -- a hack because then the substitution wouldn't be idempotent
1375 -- any more (v is an OutId). And this just just as well.
1377 zap b = b `setIdOccInfo` NoOccInfo
1379 simplCaseBinder env other_scrut case_bndr
1380 = simplBinder env case_bndr `thenSmpl` \ (env, case_bndr') ->
1381 returnSmpl (env, case_bndr', \ bndr -> bndr) -- NoOp on bndr
1387 simplAlts :: SimplEnv
1388 -> (InId -> InId) -- Occ-info zapper
1389 -> [AltCon] -- Alternatives the scrutinee can't be
1390 -- in the default case
1391 -> OutId -- Case binder
1392 -> [InAlt] -> SimplCont
1393 -> SimplM [OutAlt] -- Includes the continuation
1395 simplAlts env zap_occ_info handled_cons case_bndr' alts cont'
1396 = mapSmpl simpl_alt alts
1398 inst_tys' = tyConAppArgs (idType case_bndr')
1400 simpl_alt (DEFAULT, _, rhs)
1402 -- In the default case we record the constructors that the
1403 -- case-binder *can't* be.
1404 -- We take advantage of any OtherCon info in the case scrutinee
1405 case_bndr_w_unf = case_bndr' `setIdUnfolding` mkOtherCon handled_cons
1406 env_with_unf = modifyInScope env case_bndr' case_bndr_w_unf
1408 simplExprC env_with_unf rhs cont' `thenSmpl` \ rhs' ->
1409 returnSmpl (DEFAULT, [], rhs')
1411 simpl_alt (con, vs, rhs)
1412 = -- Deal with the pattern-bound variables
1413 -- Mark the ones that are in ! positions in the data constructor
1414 -- as certainly-evaluated.
1415 -- NB: it happens that simplBinders does *not* erase the OtherCon
1416 -- form of unfolding, so it's ok to add this info before
1417 -- doing simplBinders
1418 simplBinders env (add_evals con vs) `thenSmpl` \ (env, vs') ->
1420 -- Bind the case-binder to (con args)
1422 unfolding = mkUnfolding False (mkAltExpr con vs' inst_tys')
1423 env_with_unf = modifyInScope env case_bndr' (case_bndr' `setIdUnfolding` unfolding)
1425 simplExprC env_with_unf rhs cont' `thenSmpl` \ rhs' ->
1426 returnSmpl (con, vs', rhs')
1429 -- add_evals records the evaluated-ness of the bound variables of
1430 -- a case pattern. This is *important*. Consider
1431 -- data T = T !Int !Int
1433 -- case x of { T a b -> T (a+1) b }
1435 -- We really must record that b is already evaluated so that we don't
1436 -- go and re-evaluate it when constructing the result.
1438 add_evals (DataAlt dc) vs = cat_evals vs (dataConRepStrictness dc)
1439 add_evals other_con vs = vs
1441 cat_evals [] [] = []
1442 cat_evals (v:vs) (str:strs)
1443 | isTyVar v = v : cat_evals vs (str:strs)
1444 | isMarkedStrict str = evald_v : cat_evals vs strs
1445 | otherwise = zapped_v : cat_evals vs strs
1447 zapped_v = zap_occ_info v
1448 evald_v = zapped_v `setIdUnfolding` mkOtherCon []
1452 %************************************************************************
1454 \subsection{Known constructor}
1456 %************************************************************************
1458 We are a bit careful with occurrence info. Here's an example
1460 (\x* -> case x of (a*, b) -> f a) (h v, e)
1462 where the * means "occurs once". This effectively becomes
1463 case (h v, e) of (a*, b) -> f a)
1465 let a* = h v; b = e in f a
1469 All this should happen in one sweep.
1472 knownCon :: SimplEnv -> AltCon -> [OutExpr]
1473 -> InId -> [InAlt] -> SimplCont
1474 -> SimplM FloatsWithExpr
1476 knownCon env con args bndr alts cont
1477 = tick (KnownBranch bndr) `thenSmpl_`
1478 case findAlt con alts of
1479 (DEFAULT, bs, rhs) -> ASSERT( null bs )
1480 simplNonRecX env bndr scrut $ \ env ->
1481 -- This might give rise to a binding with non-atomic args
1482 -- like x = Node (f x) (g x)
1483 -- but no harm will be done
1484 simplExprF env rhs cont
1487 LitAlt lit -> Lit lit
1488 DataAlt dc -> mkConApp dc args
1490 (LitAlt lit, bs, rhs) -> ASSERT( null bs )
1491 simplNonRecX env bndr (Lit lit) $ \ env ->
1492 simplExprF env rhs cont
1494 (DataAlt dc, bs, rhs) -> ASSERT( length bs + n_tys == length args )
1495 bind_args env bs (drop n_tys args) $ \ env ->
1497 con_app = mkConApp dc (take n_tys args ++ con_args)
1498 con_args = [substExpr (getSubst env) (varToCoreExpr b) | b <- bs]
1499 -- args are aready OutExprs, but bs are InIds
1501 simplNonRecX env bndr con_app $ \ env ->
1502 simplExprF env rhs cont
1504 n_tys = dataConNumInstArgs dc -- Non-existential type args
1506 bind_args env [] _ thing_inside = thing_inside env
1508 bind_args env (b:bs) (Type ty : args) thing_inside
1509 = bind_args (extendSubst env b (DoneTy ty)) bs args thing_inside
1511 bind_args env (b:bs) (arg : args) thing_inside
1512 = simplNonRecX env b arg $ \ env ->
1513 bind_args env bs args thing_inside
1517 %************************************************************************
1519 \subsection{Duplicating continuations}
1521 %************************************************************************
1524 prepareCaseCont :: SimplEnv
1525 -> [InAlt] -> SimplCont
1526 -> SimplM (FloatsWith (SimplCont,SimplCont))
1527 -- Return a duplicatable continuation, a non-duplicable part
1528 -- plus some extra bindings
1530 -- No need to make it duplicatable if there's only one alternative
1531 prepareCaseCont env [alt] cont = returnSmpl (emptyFloats env, (cont, mkBoringStop (contResultType cont)))
1532 prepareCaseCont env alts cont = mkDupableCont env cont
1536 mkDupableCont :: SimplEnv -> SimplCont
1537 -> SimplM (FloatsWith (SimplCont, SimplCont))
1539 mkDupableCont env cont
1540 | contIsDupable cont
1541 = returnSmpl (emptyFloats env, (cont, mkBoringStop (contResultType cont)))
1543 mkDupableCont env (CoerceIt ty cont)
1544 = mkDupableCont env cont `thenSmpl` \ (floats, (dup_cont, nondup_cont)) ->
1545 returnSmpl (floats, (CoerceIt ty dup_cont, nondup_cont))
1547 mkDupableCont env (InlinePlease cont)
1548 = mkDupableCont env cont `thenSmpl` \ (floats, (dup_cont, nondup_cont)) ->
1549 returnSmpl (floats, (InlinePlease dup_cont, nondup_cont))
1551 mkDupableCont env cont@(ArgOf _ arg_ty _ _)
1552 = returnSmpl (emptyFloats env, (mkBoringStop arg_ty, cont))
1553 -- Do *not* duplicate an ArgOf continuation
1554 -- Because ArgOf continuations are opaque, we gain nothing by
1555 -- propagating them into the expressions, and we do lose a lot.
1556 -- Here's an example:
1557 -- && (case x of { T -> F; F -> T }) E
1558 -- Now, && is strict so we end up simplifying the case with
1559 -- an ArgOf continuation. If we let-bind it, we get
1561 -- let $j = \v -> && v E
1562 -- in simplExpr (case x of { T -> F; F -> T })
1563 -- (ArgOf (\r -> $j r)
1564 -- And after simplifying more we get
1566 -- let $j = \v -> && v E
1567 -- in case of { T -> $j F; F -> $j T }
1568 -- Which is a Very Bad Thing
1570 -- The desire not to duplicate is the entire reason that
1571 -- mkDupableCont returns a pair of continuations.
1573 -- The original plan had:
1574 -- e.g. (...strict-fn...) [...hole...]
1576 -- let $j = \a -> ...strict-fn...
1577 -- in $j [...hole...]
1579 mkDupableCont env (ApplyTo _ arg se cont)
1580 = -- e.g. [...hole...] (...arg...)
1582 -- let a = ...arg...
1583 -- in [...hole...] a
1584 simplExpr (setInScope se env) arg `thenSmpl` \ arg' ->
1586 mkDupableCont env cont `thenSmpl` \ (floats, (dup_cont, nondup_cont)) ->
1587 addFloats env floats $ \ env ->
1589 if exprIsDupable arg' then
1590 returnSmpl (emptyFloats env, (ApplyTo OkToDup arg' (zapSubstEnv se) dup_cont, nondup_cont))
1592 newId FSLIT("a") (exprType arg') `thenSmpl` \ arg_id ->
1594 tick (CaseOfCase arg_id) `thenSmpl_`
1595 -- Want to tick here so that we go round again,
1596 -- and maybe copy or inline the code.
1597 -- Not strictly CaseOfCase, but never mind
1599 returnSmpl (unitFloat env arg_id arg',
1600 (ApplyTo OkToDup (Var arg_id) (zapSubstEnv se) dup_cont,
1602 -- But what if the arg should be case-bound?
1603 -- This has been this way for a long time, so I'll leave it,
1604 -- but I can't convince myself that it's right.
1607 mkDupableCont env (Select _ case_bndr alts se cont)
1608 = -- e.g. (case [...hole...] of { pi -> ei })
1610 -- let ji = \xij -> ei
1611 -- in case [...hole...] of { pi -> ji xij }
1612 tick (CaseOfCase case_bndr) `thenSmpl_`
1614 alt_env = setInScope se env
1616 prepareCaseCont alt_env alts cont `thenSmpl` \ (floats1, (dup_cont, nondup_cont)) ->
1617 addFloats alt_env floats1 $ \ alt_env ->
1619 simplBinder alt_env case_bndr `thenSmpl` \ (alt_env, case_bndr') ->
1620 -- NB: simplBinder does not zap deadness occ-info, so
1621 -- a dead case_bndr' will still advertise its deadness
1622 -- This is really important because in
1623 -- case e of b { (# a,b #) -> ... }
1624 -- b is always dead, and indeed we are not allowed to bind b to (# a,b #),
1625 -- which might happen if e was an explicit unboxed pair and b wasn't marked dead.
1626 -- In the new alts we build, we have the new case binder, so it must retain
1629 mkDupableAlts alt_env case_bndr' alts dup_cont `thenSmpl` \ (floats2, alts') ->
1630 addFloats alt_env floats2 $ \ alt_env ->
1631 returnSmpl (emptyFloats alt_env,
1632 (Select OkToDup case_bndr' alts' (zapSubstEnv se)
1633 (mkBoringStop (contResultType dup_cont)),
1636 mkDupableAlts :: SimplEnv -> OutId -> [InAlt] -> SimplCont
1637 -> SimplM (FloatsWith [InAlt])
1638 -- Absorbs the continuation into the new alternatives
1640 mkDupableAlts env case_bndr' alts dupable_cont
1643 go env [] = returnSmpl (emptyFloats env, [])
1645 = mkDupableAlt env case_bndr' dupable_cont alt `thenSmpl` \ (floats1, alt') ->
1646 addFloats env floats1 $ \ env ->
1647 go env alts `thenSmpl` \ (floats2, alts') ->
1648 returnSmpl (floats2, alt' : alts')
1650 mkDupableAlt env case_bndr' cont alt@(con, bndrs, rhs)
1651 = simplBinders env bndrs `thenSmpl` \ (env, bndrs') ->
1652 simplExprC env rhs cont `thenSmpl` \ rhs' ->
1654 if exprIsDupable rhs' then
1655 returnSmpl (emptyFloats env, (con, bndrs', rhs'))
1656 -- It is worth checking for a small RHS because otherwise we
1657 -- get extra let bindings that may cause an extra iteration of the simplifier to
1658 -- inline back in place. Quite often the rhs is just a variable or constructor.
1659 -- The Ord instance of Maybe in PrelMaybe.lhs, for example, took several extra
1660 -- iterations because the version with the let bindings looked big, and so wasn't
1661 -- inlined, but after the join points had been inlined it looked smaller, and so
1664 -- NB: we have to check the size of rhs', not rhs.
1665 -- Duplicating a small InAlt might invalidate occurrence information
1666 -- However, if it *is* dupable, we return the *un* simplified alternative,
1667 -- because otherwise we'd need to pair it up with an empty subst-env....
1668 -- but we only have one env shared between all the alts.
1669 -- (Remember we must zap the subst-env before re-simplifying something).
1670 -- Rather than do this we simply agree to re-simplify the original (small) thing later.
1674 rhs_ty' = exprType rhs'
1675 used_bndrs' = filter (not . isDeadBinder) (case_bndr' : bndrs')
1676 -- The deadness info on the new binders is unscathed
1678 -- If we try to lift a primitive-typed something out
1679 -- for let-binding-purposes, we will *caseify* it (!),
1680 -- with potentially-disastrous strictness results. So
1681 -- instead we turn it into a function: \v -> e
1682 -- where v::State# RealWorld#. The value passed to this function
1683 -- is realworld#, which generates (almost) no code.
1685 -- There's a slight infelicity here: we pass the overall
1686 -- case_bndr to all the join points if it's used in *any* RHS,
1687 -- because we don't know its usage in each RHS separately
1689 -- We used to say "&& isUnLiftedType rhs_ty'" here, but now
1690 -- we make the join point into a function whenever used_bndrs'
1691 -- is empty. This makes the join-point more CPR friendly.
1692 -- Consider: let j = if .. then I# 3 else I# 4
1693 -- in case .. of { A -> j; B -> j; C -> ... }
1695 -- Now CPR doesn't w/w j because it's a thunk, so
1696 -- that means that the enclosing function can't w/w either,
1697 -- which is a lose. Here's the example that happened in practice:
1698 -- kgmod :: Int -> Int -> Int
1699 -- kgmod x y = if x > 0 && y < 0 || x < 0 && y > 0
1703 -- I have seen a case alternative like this:
1704 -- True -> \v -> ...
1705 -- It's a bit silly to add the realWorld dummy arg in this case, making
1708 -- (the \v alone is enough to make CPR happy) but I think it's rare
1710 ( if null used_bndrs'
1711 then newId FSLIT("w") realWorldStatePrimTy `thenSmpl` \ rw_id ->
1712 returnSmpl ([rw_id], [Var realWorldPrimId])
1714 returnSmpl (used_bndrs', map varToCoreExpr used_bndrs')
1715 ) `thenSmpl` \ (final_bndrs', final_args) ->
1717 -- See comment about "$j" name above
1718 newId (encodeFS FSLIT("$j")) (mkPiTypes final_bndrs' rhs_ty') `thenSmpl` \ join_bndr ->
1719 -- Notice the funky mkPiTypes. If the contructor has existentials
1720 -- it's possible that the join point will be abstracted over
1721 -- type varaibles as well as term variables.
1722 -- Example: Suppose we have
1723 -- data T = forall t. C [t]
1725 -- case (case e of ...) of
1726 -- C t xs::[t] -> rhs
1727 -- We get the join point
1728 -- let j :: forall t. [t] -> ...
1729 -- j = /\t \xs::[t] -> rhs
1731 -- case (case e of ...) of
1732 -- C t xs::[t] -> j t xs
1734 -- We make the lambdas into one-shot-lambdas. The
1735 -- join point is sure to be applied at most once, and doing so
1736 -- prevents the body of the join point being floated out by
1737 -- the full laziness pass
1738 really_final_bndrs = map one_shot final_bndrs'
1739 one_shot v | isId v = setOneShotLambda v
1741 join_rhs = mkLams really_final_bndrs rhs'
1742 join_call = mkApps (Var join_bndr) final_args
1744 returnSmpl (unitFloat env join_bndr join_rhs, (con, bndrs', join_call))