2 % (c) The GRASP/AQUA Project, Glasgow University, 1993-1998
4 \section[Specialise]{Stamping out overloading, and (optionally) polymorphism}
7 -- The above warning supression flag is a temporary kludge.
8 -- While working on this module you are encouraged to remove it and fix
9 -- any warnings in the module. See
10 -- http://hackage.haskell.org/trac/ghc/wiki/Commentary/CodingStyle#Warnings
13 module Specialise ( specProgram ) where
15 #include "HsVersions.h"
17 import Id ( Id, idName, idType, mkUserLocal, idCoreRules, idUnfolding,
18 idInlineActivation, setInlineActivation, setIdUnfolding,
19 isLocalId, isDataConWorkId, idArity, setIdArity )
20 import TcType ( Type, mkTyVarTy, tcSplitSigmaTy,
21 tyVarsOfTypes, tyVarsOfTheta, isClassPred,
22 tcCmpType, isUnLiftedType
24 import CoreSubst ( Subst, mkEmptySubst, extendTvSubstList, lookupIdSubst,
25 substBndr, substBndrs, substTy, substInScope,
26 cloneIdBndr, cloneIdBndrs, cloneRecIdBndrs,
29 import CoreUnfold ( mkUnfolding )
35 import CoreUtils ( exprIsTrivial, applyTypeToArgs, mkPiTypes )
36 import CoreFVs ( exprFreeVars, exprsFreeVars, idFreeVars )
37 import UniqSupply ( UniqSupply,
42 import MkId ( voidArgId, realWorldPrimId )
44 import Maybes ( catMaybes, isJust )
52 %************************************************************************
54 \subsection[notes-Specialise]{Implementation notes [SLPJ, Aug 18 1993]}
56 %************************************************************************
58 These notes describe how we implement specialisation to eliminate
61 The specialisation pass works on Core
62 syntax, complete with all the explicit dictionary application,
63 abstraction and construction as added by the type checker. The
64 existing type checker remains largely as it is.
66 One important thought: the {\em types} passed to an overloaded
67 function, and the {\em dictionaries} passed are mutually redundant.
68 If the same function is applied to the same type(s) then it is sure to
69 be applied to the same dictionary(s)---or rather to the same {\em
70 values}. (The arguments might look different but they will evaluate
73 Second important thought: we know that we can make progress by
74 treating dictionary arguments as static and worth specialising on. So
75 we can do without binding-time analysis, and instead specialise on
76 dictionary arguments and no others.
85 and suppose f is overloaded.
87 STEP 1: CALL-INSTANCE COLLECTION
89 We traverse <body>, accumulating all applications of f to types and
92 (Might there be partial applications, to just some of its types and
93 dictionaries? In principle yes, but in practice the type checker only
94 builds applications of f to all its types and dictionaries, so partial
95 applications could only arise as a result of transformation, and even
96 then I think it's unlikely. In any case, we simply don't accumulate such
97 partial applications.)
102 So now we have a collection of calls to f:
106 Notice that f may take several type arguments. To avoid ambiguity, we
107 say that f is called at type t1/t2 and t3/t4.
109 We take equivalence classes using equality of the *types* (ignoring
110 the dictionary args, which as mentioned previously are redundant).
112 STEP 3: SPECIALISATION
114 For each equivalence class, choose a representative (f t1 t2 d1 d2),
115 and create a local instance of f, defined thus:
117 f@t1/t2 = <f_rhs> t1 t2 d1 d2
119 f_rhs presumably has some big lambdas and dictionary lambdas, so lots
120 of simplification will now result. However we don't actually *do* that
121 simplification. Rather, we leave it for the simplifier to do. If we
122 *did* do it, though, we'd get more call instances from the specialised
123 RHS. We can work out what they are by instantiating the call-instance
124 set from f's RHS with the types t1, t2.
126 Add this new id to f's IdInfo, to record that f has a specialised version.
128 Before doing any of this, check that f's IdInfo doesn't already
129 tell us about an existing instance of f at the required type/s.
130 (This might happen if specialisation was applied more than once, or
131 it might arise from user SPECIALIZE pragmas.)
135 Wait a minute! What if f is recursive? Then we can't just plug in
136 its right-hand side, can we?
138 But it's ok. The type checker *always* creates non-recursive definitions
139 for overloaded recursive functions. For example:
141 f x = f (x+x) -- Yes I know its silly
145 f a (d::Num a) = let p = +.sel a d
147 letrec fl (y::a) = fl (p y y)
151 We still have recusion for non-overloaded functions which we
152 speciailise, but the recursive call should get specialised to the
153 same recursive version.
159 All this is crystal clear when the function is applied to *constant
160 types*; that is, types which have no type variables inside. But what if
161 it is applied to non-constant types? Suppose we find a call of f at type
162 t1/t2. There are two possibilities:
164 (a) The free type variables of t1, t2 are in scope at the definition point
165 of f. In this case there's no problem, we proceed just as before. A common
166 example is as follows. Here's the Haskell:
171 After typechecking we have
173 g a (d::Num a) (y::a) = let f b (d'::Num b) (x::b) = +.sel b d' x x
174 in +.sel a d (f a d y) (f a d y)
176 Notice that the call to f is at type type "a"; a non-constant type.
177 Both calls to f are at the same type, so we can specialise to give:
179 g a (d::Num a) (y::a) = let f@a (x::a) = +.sel a d x x
180 in +.sel a d (f@a y) (f@a y)
183 (b) The other case is when the type variables in the instance types
184 are *not* in scope at the definition point of f. The example we are
185 working with above is a good case. There are two instances of (+.sel a d),
186 but "a" is not in scope at the definition of +.sel. Can we do anything?
187 Yes, we can "common them up", a sort of limited common sub-expression deal.
190 g a (d::Num a) (y::a) = let +.sel@a = +.sel a d
191 f@a (x::a) = +.sel@a x x
192 in +.sel@a (f@a y) (f@a y)
194 This can save work, and can't be spotted by the type checker, because
195 the two instances of +.sel weren't originally at the same type.
199 * There are quite a few variations here. For example, the defn of
200 +.sel could be floated ouside the \y, to attempt to gain laziness.
201 It certainly mustn't be floated outside the \d because the d has to
204 * We don't want to inline f_rhs in this case, because
205 that will duplicate code. Just commoning up the call is the point.
207 * Nothing gets added to +.sel's IdInfo.
209 * Don't bother unless the equivalence class has more than one item!
211 Not clear whether this is all worth it. It is of course OK to
212 simply discard call-instances when passing a big lambda.
214 Polymorphism 2 -- Overloading
216 Consider a function whose most general type is
218 f :: forall a b. Ord a => [a] -> b -> b
220 There is really no point in making a version of g at Int/Int and another
221 at Int/Bool, because it's only instancing the type variable "a" which
222 buys us any efficiency. Since g is completely polymorphic in b there
223 ain't much point in making separate versions of g for the different
226 That suggests that we should identify which of g's type variables
227 are constrained (like "a") and which are unconstrained (like "b").
228 Then when taking equivalence classes in STEP 2, we ignore the type args
229 corresponding to unconstrained type variable. In STEP 3 we make
230 polymorphic versions. Thus:
232 f@t1/ = /\b -> <f_rhs> t1 b d1 d2
241 f a (d::Num a) = let g = ...
243 ...(let d1::Ord a = Num.Ord.sel a d in g a d1)...
245 Here, g is only called at one type, but the dictionary isn't in scope at the
246 definition point for g. Usually the type checker would build a
247 definition for d1 which enclosed g, but the transformation system
248 might have moved d1's defn inward. Solution: float dictionary bindings
249 outwards along with call instances.
253 f x = let g p q = p==q
259 Before specialisation, leaving out type abstractions we have
261 f df x = let g :: Eq a => a -> a -> Bool
263 h :: Num a => a -> a -> (a, Bool)
264 h dh r s = let deq = eqFromNum dh
265 in (+ dh r s, g deq r s)
269 After specialising h we get a specialised version of h, like this:
271 h' r s = let deq = eqFromNum df
272 in (+ df r s, g deq r s)
274 But we can't naively make an instance for g from this, because deq is not in scope
275 at the defn of g. Instead, we have to float out the (new) defn of deq
276 to widen its scope. Notice that this floating can't be done in advance -- it only
277 shows up when specialisation is done.
279 User SPECIALIZE pragmas
280 ~~~~~~~~~~~~~~~~~~~~~~~
281 Specialisation pragmas can be digested by the type checker, and implemented
282 by adding extra definitions along with that of f, in the same way as before
284 f@t1/t2 = <f_rhs> t1 t2 d1 d2
286 Indeed the pragmas *have* to be dealt with by the type checker, because
287 only it knows how to build the dictionaries d1 and d2! For example
289 g :: Ord a => [a] -> [a]
290 {-# SPECIALIZE f :: [Tree Int] -> [Tree Int] #-}
292 Here, the specialised version of g is an application of g's rhs to the
293 Ord dictionary for (Tree Int), which only the type checker can conjure
294 up. There might not even *be* one, if (Tree Int) is not an instance of
295 Ord! (All the other specialision has suitable dictionaries to hand
298 Problem. The type checker doesn't have to hand a convenient <f_rhs>, because
299 it is buried in a complex (as-yet-un-desugared) binding group.
302 f@t1/t2 = f* t1 t2 d1 d2
304 where f* is the Id f with an IdInfo which says "inline me regardless!".
305 Indeed all the specialisation could be done in this way.
306 That in turn means that the simplifier has to be prepared to inline absolutely
307 any in-scope let-bound thing.
310 Again, the pragma should permit polymorphism in unconstrained variables:
312 h :: Ord a => [a] -> b -> b
313 {-# SPECIALIZE h :: [Int] -> b -> b #-}
315 We *insist* that all overloaded type variables are specialised to ground types,
316 (and hence there can be no context inside a SPECIALIZE pragma).
317 We *permit* unconstrained type variables to be specialised to
319 - or left as a polymorphic type variable
320 but nothing in between. So
322 {-# SPECIALIZE h :: [Int] -> [c] -> [c] #-}
324 is *illegal*. (It can be handled, but it adds complication, and gains the
328 SPECIALISING INSTANCE DECLARATIONS
329 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
332 instance Foo a => Foo [a] where
334 {-# SPECIALIZE instance Foo [Int] #-}
336 The original instance decl creates a dictionary-function
339 dfun.Foo.List :: forall a. Foo a -> Foo [a]
341 The SPECIALIZE pragma just makes a specialised copy, just as for
342 ordinary function definitions:
344 dfun.Foo.List@Int :: Foo [Int]
345 dfun.Foo.List@Int = dfun.Foo.List Int dFooInt
347 The information about what instance of the dfun exist gets added to
348 the dfun's IdInfo in the same way as a user-defined function too.
351 Automatic instance decl specialisation?
352 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
353 Can instance decls be specialised automatically? It's tricky.
354 We could collect call-instance information for each dfun, but
355 then when we specialised their bodies we'd get new call-instances
356 for ordinary functions; and when we specialised their bodies, we might get
357 new call-instances of the dfuns, and so on. This all arises because of
358 the unrestricted mutual recursion between instance decls and value decls.
360 Still, there's no actual problem; it just means that we may not do all
361 the specialisation we could theoretically do.
363 Furthermore, instance decls are usually exported and used non-locally,
364 so we'll want to compile enough to get those specialisations done.
366 Lastly, there's no such thing as a local instance decl, so we can
367 survive solely by spitting out *usage* information, and then reading that
368 back in as a pragma when next compiling the file. So for now,
369 we only specialise instance decls in response to pragmas.
372 SPITTING OUT USAGE INFORMATION
373 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
375 To spit out usage information we need to traverse the code collecting
376 call-instance information for all imported (non-prelude?) functions
377 and data types. Then we equivalence-class it and spit it out.
379 This is done at the top-level when all the call instances which escape
380 must be for imported functions and data types.
382 *** Not currently done ***
385 Partial specialisation by pragmas
386 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
387 What about partial specialisation:
389 k :: (Ord a, Eq b) => [a] -> b -> b -> [a]
390 {-# SPECIALIZE k :: Eq b => [Int] -> b -> b -> [a] #-}
394 {-# SPECIALIZE k :: Eq b => [Int] -> [b] -> [b] -> [a] #-}
396 Seems quite reasonable. Similar things could be done with instance decls:
398 instance (Foo a, Foo b) => Foo (a,b) where
400 {-# SPECIALIZE instance Foo a => Foo (a,Int) #-}
401 {-# SPECIALIZE instance Foo b => Foo (Int,b) #-}
403 Ho hum. Things are complex enough without this. I pass.
406 Requirements for the simplifer
407 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
408 The simplifier has to be able to take advantage of the specialisation.
410 * When the simplifier finds an application of a polymorphic f, it looks in
411 f's IdInfo in case there is a suitable instance to call instead. This converts
413 f t1 t2 d1 d2 ===> f_t1_t2
415 Note that the dictionaries get eaten up too!
417 * Dictionary selection operations on constant dictionaries must be
420 +.sel Int d ===> +Int
422 The obvious way to do this is in the same way as other specialised
423 calls: +.sel has inside it some IdInfo which tells that if it's applied
424 to the type Int then it should eat a dictionary and transform to +Int.
426 In short, dictionary selectors need IdInfo inside them for constant
429 * Exactly the same applies if a superclass dictionary is being
432 Eq.sel Int d ===> dEqInt
434 * Something similar applies to dictionary construction too. Suppose
435 dfun.Eq.List is the function taking a dictionary for (Eq a) to
436 one for (Eq [a]). Then we want
438 dfun.Eq.List Int d ===> dEq.List_Int
440 Where does the Eq [Int] dictionary come from? It is built in
441 response to a SPECIALIZE pragma on the Eq [a] instance decl.
443 In short, dfun Ids need IdInfo with a specialisation for each
444 constant instance of their instance declaration.
446 All this uses a single mechanism: the SpecEnv inside an Id
449 What does the specialisation IdInfo look like?
450 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
452 The SpecEnv of an Id maps a list of types (the template) to an expression
456 For example, if f has this SpecInfo:
458 [Int, a] -> \d:Ord Int. f' a
460 it means that we can replace the call
462 f Int t ===> (\d. f' t)
464 This chucks one dictionary away and proceeds with the
465 specialised version of f, namely f'.
468 What can't be done this way?
469 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
470 There is no way, post-typechecker, to get a dictionary for (say)
471 Eq a from a dictionary for Eq [a]. So if we find
475 we can't transform to
480 eqList :: (a->a->Bool) -> [a] -> [a] -> Bool
482 Of course, we currently have no way to automatically derive
483 eqList, nor to connect it to the Eq [a] instance decl, but you
484 can imagine that it might somehow be possible. Taking advantage
485 of this is permanently ruled out.
487 Still, this is no great hardship, because we intend to eliminate
488 overloading altogether anyway!
490 A note about non-tyvar dictionaries
491 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
492 Some Ids have types like
494 forall a,b,c. Eq a -> Ord [a] -> tau
496 This seems curious at first, because we usually only have dictionary
497 args whose types are of the form (C a) where a is a type variable.
498 But this doesn't hold for the functions arising from instance decls,
499 which sometimes get arguements with types of form (C (T a)) for some
502 Should we specialise wrt this compound-type dictionary? We used to say
504 "This is a heuristic judgement, as indeed is the fact that we
505 specialise wrt only dictionaries. We choose *not* to specialise
506 wrt compound dictionaries because at the moment the only place
507 they show up is in instance decls, where they are simply plugged
508 into a returned dictionary. So nothing is gained by specialising
511 But it is simpler and more uniform to specialise wrt these dicts too;
512 and in future GHC is likely to support full fledged type signatures
514 f :: Eq [(a,b)] => ...
517 %************************************************************************
519 \subsubsection{The new specialiser}
521 %************************************************************************
523 Our basic game plan is this. For let(rec) bound function
524 f :: (C a, D c) => (a,b,c,d) -> Bool
526 * Find any specialised calls of f, (f ts ds), where
527 ts are the type arguments t1 .. t4, and
528 ds are the dictionary arguments d1 .. d2.
530 * Add a new definition for f1 (say):
532 f1 = /\ b d -> (..body of f..) t1 b t3 d d1 d2
534 Note that we abstract over the unconstrained type arguments.
538 [t1,b,t3,d] |-> \d1 d2 -> f1 b d
540 to the specialisations of f. This will be used by the
541 simplifier to replace calls
542 (f t1 t2 t3 t4) da db
544 (\d1 d1 -> f1 t2 t4) da db
546 All the stuff about how many dictionaries to discard, and what types
547 to apply the specialised function to, are handled by the fact that the
548 SpecEnv contains a template for the result of the specialisation.
550 We don't build *partial* specialisations for f. For example:
552 f :: Eq a => a -> a -> Bool
553 {-# SPECIALISE f :: (Eq b, Eq c) => (b,c) -> (b,c) -> Bool #-}
555 Here, little is gained by making a specialised copy of f.
556 There's a distinct danger that the specialised version would
557 first build a dictionary for (Eq b, Eq c), and then select the (==)
558 method from it! Even if it didn't, not a great deal is saved.
560 We do, however, generate polymorphic, but not overloaded, specialisations:
562 f :: Eq a => [a] -> b -> b -> b
563 {#- SPECIALISE f :: [Int] -> b -> b -> b #-}
565 Hence, the invariant is this:
567 *** no specialised version is overloaded ***
570 %************************************************************************
572 \subsubsection{The exported function}
574 %************************************************************************
577 specProgram :: UniqSupply -> [CoreBind] -> [CoreBind]
578 specProgram us binds = initSM us (do (binds', uds') <- go binds
579 return (dumpAllDictBinds uds' binds'))
581 -- We need to start with a Subst that knows all the things
582 -- that are in scope, so that the substitution engine doesn't
583 -- accidentally re-use a unique that's already in use
584 -- Easiest thing is to do it all at once, as if all the top-level
585 -- decls were mutually recursive
586 top_subst = mkEmptySubst (mkInScopeSet (mkVarSet (bindersOfBinds binds)))
588 go [] = return ([], emptyUDs)
589 go (bind:binds) = do (binds', uds) <- go binds
590 (bind', uds') <- specBind top_subst bind uds
591 return (bind' ++ binds', uds')
594 %************************************************************************
596 \subsubsection{@specExpr@: the main function}
598 %************************************************************************
601 specVar :: Subst -> Id -> CoreExpr
602 specVar subst v = lookupIdSubst subst v
604 specExpr :: Subst -> CoreExpr -> SpecM (CoreExpr, UsageDetails)
605 -- We carry a substitution down:
606 -- a) we must clone any binding that might float outwards,
607 -- to avoid name clashes
608 -- b) we carry a type substitution to use when analysing
609 -- the RHS of specialised bindings (no type-let!)
611 ---------------- First the easy cases --------------------
612 specExpr subst (Type ty) = return (Type (substTy subst ty), emptyUDs)
613 specExpr subst (Var v) = return (specVar subst v, emptyUDs)
614 specExpr _ (Lit lit) = return (Lit lit, emptyUDs)
615 specExpr subst (Cast e co) = do
616 (e', uds) <- specExpr subst e
617 return ((Cast e' (substTy subst co)), uds)
618 specExpr subst (Note note body) = do
619 (body', uds) <- specExpr subst body
620 return (Note (specNote subst note) body', uds)
623 ---------------- Applications might generate a call instance --------------------
624 specExpr subst expr@(App {})
627 go (App fun arg) args = do (arg', uds_arg) <- specExpr subst arg
628 (fun', uds_app) <- go fun (arg':args)
629 return (App fun' arg', uds_arg `plusUDs` uds_app)
631 go (Var f) args = case specVar subst f of
632 Var f' -> return (Var f', mkCallUDs f' args)
633 e' -> return (e', emptyUDs) -- I don't expect this!
634 go other _ = specExpr subst other
636 ---------------- Lambda/case require dumping of usage details --------------------
637 specExpr subst e@(Lam _ _) = do
638 (body', uds) <- specExpr subst' body
639 let (filtered_uds, body'') = dumpUDs bndrs' uds body'
640 return (mkLams bndrs' body'', filtered_uds)
642 (bndrs, body) = collectBinders e
643 (subst', bndrs') = substBndrs subst bndrs
644 -- More efficient to collect a group of binders together all at once
645 -- and we don't want to split a lambda group with dumped bindings
647 specExpr subst (Case scrut case_bndr ty alts) = do
648 (scrut', uds_scrut) <- specExpr subst scrut
649 (alts', uds_alts) <- mapAndCombineSM spec_alt alts
650 return (Case scrut' case_bndr' (substTy subst ty) alts', uds_scrut `plusUDs` uds_alts)
652 (subst_alt, case_bndr') = substBndr subst case_bndr
653 -- No need to clone case binder; it can't float like a let(rec)
655 spec_alt (con, args, rhs) = do
656 (rhs', uds) <- specExpr subst_rhs rhs
657 let (uds', rhs'') = dumpUDs args uds rhs'
658 return ((con, args', rhs''), uds')
660 (subst_rhs, args') = substBndrs subst_alt args
662 ---------------- Finally, let is the interesting case --------------------
663 specExpr subst (Let bind body) = do
665 (rhs_subst, body_subst, bind') <- cloneBindSM subst bind
667 -- Deal with the body
668 (body', body_uds) <- specExpr body_subst body
670 -- Deal with the bindings
671 (binds', uds) <- specBind rhs_subst bind' body_uds
674 return (foldr Let body' binds', uds)
676 -- Must apply the type substitution to coerceions
677 specNote :: Subst -> Note -> Note
678 specNote _ note = note
681 %************************************************************************
683 \subsubsection{Dealing with a binding}
685 %************************************************************************
688 specBind :: Subst -- Use this for RHSs
690 -> UsageDetails -- Info on how the scope of the binding
691 -> SpecM ([CoreBind], -- New bindings
692 UsageDetails) -- And info to pass upstream
694 specBind rhs_subst bind body_uds
695 = do { (bind', bind_uds) <- specBindItself rhs_subst bind (calls body_uds)
696 ; return (finishSpecBind bind' bind_uds body_uds) }
698 finishSpecBind :: CoreBind -> UsageDetails -> UsageDetails -> ([CoreBind], UsageDetails)
700 (MkUD { dict_binds = rhs_dbs, calls = rhs_calls, ud_fvs = rhs_fvs })
701 (MkUD { dict_binds = body_dbs, calls = body_calls, ud_fvs = body_fvs })
702 | not (mkVarSet bndrs `intersectsVarSet` all_fvs)
703 -- Common case 1: the bound variables are not
704 -- mentioned in the dictionary bindings
705 = ([bind], MkUD { dict_binds = body_dbs `unionBags` rhs_dbs
706 -- It's important that the `unionBags` is this way round,
707 -- because body_uds may bind dictionaries that are
708 -- used in the calls passed to specDefn. So the
709 -- dictionary bindings in rhs_uds may mention
710 -- dictionaries bound in body_uds.
712 , ud_fvs = all_fvs })
714 | case bind of { NonRec {} -> True; Rec {} -> False }
715 -- Common case 2: no specialisation happened, and binding
716 -- is non-recursive. But the binding may be
717 -- mentioned in body_dbs, so we should put it first
718 = ([], MkUD { dict_binds = rhs_dbs `unionBags` ((bind, b_fvs) `consBag` body_dbs)
720 , ud_fvs = all_fvs `unionVarSet` b_fvs })
722 | otherwise -- General case: make a huge Rec (sigh)
723 = ([], MkUD { dict_binds = unitBag (Rec all_db_prs, all_db_fvs)
725 , ud_fvs = all_fvs `unionVarSet` b_fvs })
727 all_fvs = rhs_fvs `unionVarSet` body_fvs
728 all_calls = zapCalls bndrs (rhs_calls `unionCalls` body_calls)
730 bndrs = bindersOf bind
731 b_fvs = bind_fvs bind
733 (all_db_prs, all_db_fvs) = add (bind, b_fvs) $
734 foldrBag add ([], emptyVarSet) $
735 rhs_dbs `unionBags` body_dbs
736 add (NonRec b r, b_fvs) (prs, fvs) = ((b,r) : prs, b_fvs `unionVarSet` fvs)
737 add (Rec b_prs, b_fvs) (prs, fvs) = (b_prs ++ prs, b_fvs `unionVarSet` fvs)
739 ---------------------------
740 specBindItself :: Subst -> CoreBind -> CallDetails -> SpecM (CoreBind, UsageDetails)
742 -- specBindItself deals with the RHS, specialising it according
743 -- to the calls found in the body (if any)
744 specBindItself rhs_subst (NonRec fn rhs) call_info
745 = do { (rhs', rhs_uds) <- specExpr rhs_subst rhs -- Do RHS of original fn
746 ; (fn', spec_defns, spec_uds) <- specDefn rhs_subst call_info fn rhs
747 ; if null spec_defns then
748 return (NonRec fn rhs', rhs_uds)
750 return (Rec ((fn',rhs') : spec_defns), rhs_uds `plusUDs` spec_uds) }
751 -- bndr' mentions the spec_defns in its SpecEnv
752 -- Not sure why we couln't just put the spec_defns first
754 specBindItself rhs_subst (Rec pairs) call_info
755 -- Note [Specialising a recursive group]
756 = do { let (bndrs,rhss) = unzip pairs
757 ; (rhss', rhs_uds) <- mapAndCombineSM (specExpr rhs_subst) rhss
758 ; let all_calls = call_info `unionCalls` calls rhs_uds
759 ; (bndrs1, spec_defns1, spec_uds1) <- specDefns rhs_subst all_calls pairs
761 ; if null spec_defns1 then -- Common case: no specialisation
762 return (Rec (bndrs `zip` rhss'), rhs_uds)
763 else do -- Specialisation occurred; do it again
764 { (bndrs2, spec_defns2, spec_uds2) <-
765 -- pprTrace "specB" (ppr bndrs $$ ppr rhs_uds) $
766 specDefns rhs_subst (calls spec_uds1) (bndrs1 `zip` rhss)
768 ; let all_defns = spec_defns1 ++ spec_defns2 ++ zip bndrs2 rhss'
770 ; return (Rec all_defns, rhs_uds `plusUDs` spec_uds1 `plusUDs` spec_uds2) } }
773 ---------------------------
775 -> CallDetails -- Info on how it is used in its scope
776 -> [(Id,CoreExpr)] -- The things being bound and their un-processed RHS
777 -> SpecM ([Id], -- Original Ids with RULES added
778 [(Id,CoreExpr)], -- Extra, specialised bindings
779 UsageDetails) -- Stuff to fling upwards from the specialised versions
781 -- Specialise a list of bindings (the contents of a Rec), but flowing usages
782 -- upwards binding by binding. Example: { f = ...g ...; g = ...f .... }
783 -- Then if the input CallDetails has a specialised call for 'g', whose specialisation
784 -- in turn generates a specialised call for 'f', we catch that in this one sweep.
785 -- But not vice versa (it's a fixpoint problem).
787 specDefns _subst _call_info []
788 = return ([], [], emptyUDs)
789 specDefns subst call_info ((bndr,rhs):pairs)
790 = do { (bndrs', spec_defns, spec_uds) <- specDefns subst call_info pairs
791 ; let all_calls = call_info `unionCalls` calls spec_uds
792 ; (bndr', spec_defns1, spec_uds1) <- specDefn subst all_calls bndr rhs
793 ; return (bndr' : bndrs',
794 spec_defns1 ++ spec_defns,
795 spec_uds1 `plusUDs` spec_uds) }
797 ---------------------------
799 -> CallDetails -- Info on how it is used in its scope
800 -> Id -> CoreExpr -- The thing being bound and its un-processed RHS
801 -> SpecM (Id, -- Original Id with added RULES
802 [(Id,CoreExpr)], -- Extra, specialised bindings
803 UsageDetails) -- Stuff to fling upwards from the specialised versions
805 specDefn subst calls fn rhs
806 -- The first case is the interesting one
807 | rhs_tyvars `lengthIs` n_tyvars -- Rhs of fn's defn has right number of big lambdas
808 && rhs_ids `lengthAtLeast` n_dicts -- and enough dict args
809 && notNull calls_for_me -- And there are some calls to specialise
811 -- && not (certainlyWillInline (idUnfolding fn)) -- And it's not small
812 -- See Note [Inline specialisation] for why we do not
813 -- switch off specialisation for inline functions
815 = do { -- Make a specialised version for each call in calls_for_me
816 stuff <- mapM spec_call calls_for_me
817 ; let (spec_defns, spec_uds, spec_rules) = unzip3 (catMaybes stuff)
818 fn' = addIdSpecialisations fn spec_rules
819 ; return (fn', spec_defns, plusUDList spec_uds) }
821 | otherwise -- No calls or RHS doesn't fit our preconceptions
822 = WARN( notNull calls_for_me, ptext (sLit "Missed specialisation opportunity for") <+> ppr fn )
823 -- Note [Specialisation shape]
824 return (fn, [], emptyUDs)
828 fn_arity = idArity fn
829 (tyvars, theta, _) = tcSplitSigmaTy fn_type
830 n_tyvars = length tyvars
831 n_dicts = length theta
832 inline_act = idInlineActivation fn
834 -- It's important that we "see past" any INLINE pragma
835 -- else we'll fail to specialise an INLINE thing
836 (inline_rhs, rhs_inside) = dropInline rhs
837 (rhs_tyvars, rhs_ids, rhs_body) = collectTyAndValBinders rhs_inside
839 rhs_dict_ids = take n_dicts rhs_ids
840 body = mkLams (drop n_dicts rhs_ids) rhs_body
841 -- Glue back on the non-dict lambdas
843 calls_for_me = case lookupFM calls fn of
845 Just cs -> fmToList cs
847 already_covered :: [CoreExpr] -> Bool
848 already_covered args -- Note [Specialisations already covered]
849 = isJust (lookupRule (const True) (substInScope subst)
850 fn args (idCoreRules fn))
852 mk_ty_args :: [Maybe Type] -> [CoreExpr]
853 mk_ty_args call_ts = zipWithEqual "spec_call" mk_ty_arg rhs_tyvars call_ts
855 mk_ty_arg rhs_tyvar Nothing = Type (mkTyVarTy rhs_tyvar)
856 mk_ty_arg _ (Just ty) = Type ty
858 ----------------------------------------------------------
859 -- Specialise to one particular call pattern
860 spec_call :: (CallKey, ([DictExpr], VarSet)) -- Call instance
861 -> SpecM (Maybe ((Id,CoreExpr), -- Specialised definition
862 UsageDetails, -- Usage details from specialised body
863 CoreRule)) -- Info for the Id's SpecEnv
864 spec_call (CallKey call_ts, (call_ds, _))
865 = ASSERT( call_ts `lengthIs` n_tyvars && call_ds `lengthIs` n_dicts )
867 -- Suppose f's defn is f = /\ a b c -> \ d1 d2 -> rhs
868 -- Supppose the call is for f [Just t1, Nothing, Just t3] [dx1, dx2]
870 -- Construct the new binding
871 -- f1 = SUBST[a->t1,c->t3, d1->d1', d2->d2'] (/\ b -> rhs)
872 -- PLUS the usage-details
873 -- { d1' = dx1; d2' = dx2 }
874 -- where d1', d2' are cloned versions of d1,d2, with the type substitution
875 -- applied. These auxiliary bindings just avoid duplication of dx1, dx2
877 -- Note that the substitution is applied to the whole thing.
878 -- This is convenient, but just slightly fragile. Notably:
879 -- * There had better be no name clashes in a/b/c
881 -- poly_tyvars = [b] in the example above
882 -- spec_tyvars = [a,c]
883 -- ty_args = [t1,b,t3]
884 poly_tyvars = [tv | (tv, Nothing) <- rhs_tyvars `zip` call_ts]
885 spec_tv_binds = [(tv,ty) | (tv, Just ty) <- rhs_tyvars `zip` call_ts]
886 spec_ty_args = map snd spec_tv_binds
887 ty_args = mk_ty_args call_ts
888 rhs_subst = extendTvSubstList subst spec_tv_binds
890 ; (rhs_subst1, inst_dict_ids) <- cloneDictBndrs rhs_subst rhs_dict_ids
891 -- Clone rhs_dicts, including instantiating their types
893 ; let (rhs_subst2, dx_binds) = bindAuxiliaryDicts rhs_subst1 $
894 (my_zipEqual rhs_dict_ids inst_dict_ids call_ds)
895 inst_args = ty_args ++ map Var inst_dict_ids
897 ; if already_covered inst_args then
900 { -- Figure out the type of the specialised function
901 let body_ty = applyTypeToArgs rhs fn_type inst_args
902 (lam_args, app_args) -- Add a dummy argument if body_ty is unlifted
903 | isUnLiftedType body_ty -- C.f. WwLib.mkWorkerArgs
904 = (poly_tyvars ++ [voidArgId], poly_tyvars ++ [realWorldPrimId])
905 | otherwise = (poly_tyvars, poly_tyvars)
906 spec_id_ty = mkPiTypes lam_args body_ty
908 ; spec_f <- newSpecIdSM fn spec_id_ty
909 ; let spec_f_w_arity = setIdArity spec_f (max 0 (fn_arity - n_dicts))
910 -- Adding arity information just propagates it a bit faster
911 -- See Note [Arity decrease] in Simplify
913 ; (spec_rhs, rhs_uds) <- specExpr rhs_subst2 (mkLams lam_args body)
915 -- The rule to put in the function's specialisation is:
916 -- forall b, d1',d2'. f t1 b t3 d1' d2' = f1 b
917 rule_name = mkFastString ("SPEC " ++ showSDoc (ppr fn <+> ppr spec_ty_args))
918 spec_env_rule = mkLocalRule
920 inline_act -- Note [Auto-specialisation and RULES]
922 (poly_tyvars ++ inst_dict_ids)
924 (mkVarApps (Var spec_f_w_arity) app_args)
926 -- Add the { d1' = dx1; d2' = dx2 } usage stuff
927 final_uds = foldr addDictBind rhs_uds dx_binds
929 spec_pr | inline_rhs = (spec_f_w_arity `setInlineActivation` inline_act, Note InlineMe spec_rhs)
930 | otherwise = (spec_f_w_arity, spec_rhs)
932 ; return (Just (spec_pr, final_uds, spec_env_rule)) } }
935 | debugIsOn && not (equalLength xs ys && equalLength ys zs)
936 = pprPanic "my_zipEqual" (vcat [ ppr xs, ppr ys
937 , ppr fn <+> ppr call_ts
938 , ppr (idType fn), ppr theta
939 , ppr n_dicts, ppr rhs_dict_ids
941 | otherwise = zip3 xs ys zs
945 -> [(DictId,DictId,CoreExpr)] -- (orig_dict, inst_dict, dx)
946 -> (Subst, -- Substitute for all orig_dicts
947 [(DictId, CoreExpr)]) -- Auxiliary bindings
948 -- Bind any dictionary arguments to fresh names, to preserve sharing
949 -- Substitution already substitutes orig_dict -> inst_dict
950 bindAuxiliaryDicts subst triples = go subst [] triples
952 go subst binds [] = (subst, binds)
953 go subst binds ((d, dx_id, dx) : pairs)
954 | exprIsTrivial dx = go (extendIdSubst subst d dx) binds pairs
955 -- No auxiliary binding necessary
956 | otherwise = go subst_w_unf ((dx_id,dx) : binds) pairs
958 dx_id1 = dx_id `setIdUnfolding` mkUnfolding False dx
959 subst_w_unf = extendIdSubst subst d (Var dx_id1)
960 -- Important! We're going to substitute dx_id1 for d
961 -- and we want it to look "interesting", else we won't gather *any*
962 -- consequential calls. E.g.
964 -- If we specialise f for a call (f (dfun dNumInt)), we'll get
965 -- a consequent call (g d') with an auxiliary definition
967 -- We want that consequent call to look interesting
970 Note [Specialising a recursive group]
971 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
973 let rec { f x = ...g x'...
974 ; g y = ...f y'.... }
976 Here we specialise 'f' at Char; but that is very likely to lead to
977 a specialisation of 'g' at Char. We must do the latter, else the
978 whole point of specialisation is lost.
980 But we do not want to keep iterating to a fixpoint, because in the
981 presence of polymorphic recursion we might generate an infinite number
984 So we use the following heuristic:
985 * Arrange the rec block in dependency order, so far as possible
986 (the occurrence analyser already does this)
988 * Specialise it much like a sequence of lets
990 * Then go through the block a second time, feeding call-info from
991 the RHSs back in the bottom, as it were
993 In effect, the ordering maxmimises the effectiveness of each sweep,
994 and we do just two sweeps. This should catch almost every case of
995 monomorphic recursion -- the exception could be a very knotted-up
996 recursion with multiple cycles tied up together.
998 This plan is implemented in the Rec case of specBindItself.
1000 Note [Specialisations already covered]
1001 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1002 We obviously don't want to generate two specialisations for the same
1003 argument pattern. There are two wrinkles
1005 1. We do the already-covered test in specDefn, not when we generate
1006 the CallInfo in mkCallUDs. We used to test in the latter place, but
1007 we now iterate the specialiser somewhat, and the Id at the call site
1008 might therefore not have all the RULES that we can see in specDefn
1010 2. What about two specialisations where the second is an *instance*
1011 of the first? If the more specific one shows up first, we'll generate
1012 specialisations for both. If the *less* specific one shows up first,
1013 we *don't* currently generate a specialisation for the more specific
1014 one. (See the call to lookupRule in already_covered.) Reasons:
1015 (a) lookupRule doesn't say which matches are exact (bad reason)
1016 (b) if the earlier specialisation is user-provided, it's
1017 far from clear that we should auto-specialise further
1019 Note [Auto-specialisation and RULES]
1020 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1022 g :: Num a => a -> a
1025 f :: (Int -> Int) -> Int
1027 {-# RULE f g = 0 #-}
1029 Suppose that auto-specialisation makes a specialised version of
1030 g::Int->Int That version won't appear in the LHS of the RULE for f.
1031 So if the specialisation rule fires too early, the rule for f may
1034 It might be possible to add new rules, to "complete" the rewrite system.
1036 RULE forall d. g Int d = g_spec
1040 But that's a bit complicated. For now we ask the programmer's help,
1041 by *copying the INLINE activation pragma* to the auto-specialised rule.
1042 So if g says {-# NOINLINE[2] g #-}, then the auto-spec rule will also
1043 not be active until phase 2.
1046 Note [Specialisation shape]
1047 ~~~~~~~~~~~~~~~~~~~~~~~~~~~
1048 We only specialise a function if it has visible top-level lambdas
1049 corresponding to its overloading. E.g. if
1050 f :: forall a. Eq a => ....
1051 then its body must look like
1054 Reason: when specialising the body for a call (f ty dexp), we want to
1055 substitute dexp for d, and pick up specialised calls in the body of f.
1057 This doesn't always work. One example I came across was this:
1058 newtype Gen a = MkGen{ unGen :: Int -> a }
1060 choose :: Eq a => a -> Gen a
1061 choose n = MkGen (\r -> n)
1063 oneof = choose (1::Int)
1065 It's a silly exapmle, but we get
1066 choose = /\a. g `cast` co
1067 where choose doesn't have any dict arguments. Thus far I have not
1068 tried to fix this (wait till there's a real example).
1071 Note [Inline specialisations]
1072 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1073 We transfer to the specialised function any INLINE stuff from the
1074 original. This means (a) the Activation in the IdInfo, and (b) any
1075 InlineMe on the RHS. We do not, however, transfer the RuleMatchInfo
1076 since we do not expect the specialisation to occur in rewrite rules.
1078 This is a change (Jun06). Previously the idea is that the point of
1079 inlining was precisely to specialise the function at its call site,
1080 and that's not so important for the specialised copies. But
1081 *pragma-directed* specialisation now takes place in the
1082 typechecker/desugarer, with manually specified INLINEs. The
1083 specialiation here is automatic. It'd be very odd if a function
1084 marked INLINE was specialised (because of some local use), and then
1085 forever after (including importing modules) the specialised version
1086 wasn't INLINEd. After all, the programmer said INLINE!
1088 You might wonder why we don't just not specialise INLINE functions.
1089 It's because even INLINE functions are sometimes not inlined, when
1090 they aren't applied to interesting arguments. But perhaps the type
1091 arguments alone are enough to specialise (even though the args are too
1092 boring to trigger inlining), and it's certainly better to call the
1093 specialised version.
1095 A case in point is dictionary functions, which are current marked
1096 INLINE, but which are worth specialising.
1099 dropInline :: CoreExpr -> (Bool, CoreExpr)
1100 dropInline (Note InlineMe rhs) = (True, rhs)
1101 dropInline rhs = (False, rhs)
1104 %************************************************************************
1106 \subsubsection{UsageDetails and suchlike}
1108 %************************************************************************
1113 dict_binds :: !(Bag DictBind),
1114 -- Floated dictionary bindings
1115 -- The order is important;
1116 -- in ds1 `union` ds2, bindings in ds2 can depend on those in ds1
1117 -- (Remember, Bags preserve order in GHC.)
1119 calls :: !CallDetails,
1121 ud_fvs :: !VarSet -- A superset of the variables mentioned in
1122 -- either dict_binds or calls
1125 instance Outputable UsageDetails where
1126 ppr (MkUD { dict_binds = dbs, calls = calls, ud_fvs = fvs })
1127 = ptext (sLit "MkUD") <+> braces (sep (punctuate comma
1128 [ptext (sLit "binds") <+> equals <+> ppr dbs,
1129 ptext (sLit "calls") <+> equals <+> ppr calls,
1130 ptext (sLit "fvs") <+> equals <+> ppr fvs]))
1132 type DictBind = (CoreBind, VarSet)
1133 -- The set is the free vars of the binding
1134 -- both tyvars and dicts
1136 type DictExpr = CoreExpr
1138 emptyUDs :: UsageDetails
1139 emptyUDs = MkUD { dict_binds = emptyBag, calls = emptyFM, ud_fvs = emptyVarSet }
1141 ------------------------------------------------------------
1142 type CallDetails = FiniteMap Id CallInfo
1143 newtype CallKey = CallKey [Maybe Type] -- Nothing => unconstrained type argument
1145 -- CallInfo uses a FiniteMap, thereby ensuring that
1146 -- we record only one call instance for any key
1148 -- The list of types and dictionaries is guaranteed to
1149 -- match the type of f
1150 type CallInfo = FiniteMap CallKey ([DictExpr], VarSet)
1151 -- Range is dict args and the vars of the whole
1152 -- call (including tyvars)
1153 -- [*not* include the main id itself, of course]
1155 instance Outputable CallKey where
1156 ppr (CallKey ts) = ppr ts
1158 -- Type isn't an instance of Ord, so that we can control which
1159 -- instance we use. That's tiresome here. Oh well
1160 instance Eq CallKey where
1161 k1 == k2 = case k1 `compare` k2 of { EQ -> True; _ -> False }
1163 instance Ord CallKey where
1164 compare (CallKey k1) (CallKey k2) = cmpList cmp k1 k2
1166 cmp Nothing Nothing = EQ
1167 cmp Nothing (Just _) = LT
1168 cmp (Just _) Nothing = GT
1169 cmp (Just t1) (Just t2) = tcCmpType t1 t2
1171 unionCalls :: CallDetails -> CallDetails -> CallDetails
1172 unionCalls c1 c2 = plusFM_C plusFM c1 c2
1174 singleCall :: Id -> [Maybe Type] -> [DictExpr] -> UsageDetails
1175 singleCall id tys dicts
1176 = MkUD {dict_binds = emptyBag,
1177 calls = unitFM id (unitFM (CallKey tys) (dicts, call_fvs)),
1180 call_fvs = exprsFreeVars dicts `unionVarSet` tys_fvs
1181 tys_fvs = tyVarsOfTypes (catMaybes tys)
1182 -- The type args (tys) are guaranteed to be part of the dictionary
1183 -- types, because they are just the constrained types,
1184 -- and the dictionary is therefore sure to be bound
1185 -- inside the binding for any type variables free in the type;
1186 -- hence it's safe to neglect tyvars free in tys when making
1187 -- the free-var set for this call
1188 -- BUT I don't trust this reasoning; play safe and include tys_fvs
1190 -- We don't include the 'id' itself.
1192 mkCallUDs :: Id -> [CoreExpr] -> UsageDetails
1194 | not (isLocalId f) -- Imported from elsewhere
1195 || null theta -- Not overloaded
1196 || not (all isClassPred theta)
1197 -- Only specialise if all overloading is on class params.
1198 -- In ptic, with implicit params, the type args
1199 -- *don't* say what the value of the implicit param is!
1200 || not (spec_tys `lengthIs` n_tyvars)
1201 || not ( dicts `lengthIs` n_dicts)
1202 || not (any interestingDict dicts) -- Note [Interesting dictionary arguments]
1203 -- See also Note [Specialisations already covered]
1204 = -- pprTrace "mkCallUDs: discarding" (vcat [ppr f, ppr args, ppr n_tyvars, ppr n_dicts, ppr (map interestingDict dicts)])
1205 emptyUDs -- Not overloaded, or no specialisation wanted
1208 = -- pprTrace "mkCallUDs: keeping" (vcat [ppr f, ppr args, ppr n_tyvars, ppr n_dicts, ppr (map interestingDict dicts)])
1209 singleCall f spec_tys dicts
1211 (tyvars, theta, _) = tcSplitSigmaTy (idType f)
1212 constrained_tyvars = tyVarsOfTheta theta
1213 n_tyvars = length tyvars
1214 n_dicts = length theta
1216 spec_tys = [mk_spec_ty tv ty | (tv, Type ty) <- tyvars `zip` args]
1217 dicts = [dict_expr | (_, dict_expr) <- theta `zip` (drop n_tyvars args)]
1220 | tyvar `elemVarSet` constrained_tyvars = Just ty
1221 | otherwise = Nothing
1224 Note [Interesting dictionary arguments]
1225 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1227 \a.\d:Eq a. let f = ... in ...(f d)...
1228 There really is not much point in specialising f wrt the dictionary d,
1229 because the code for the specialised f is not improved at all, because
1230 d is lambda-bound. We simply get junk specialisations.
1232 What is "interesting"? Just that it has *some* structure.
1235 interestingDict :: CoreExpr -> Bool
1236 -- A dictionary argument is interesting if it has *some* structure
1237 interestingDict (Var v) = hasSomeUnfolding (idUnfolding v)
1238 || isDataConWorkId v
1239 interestingDict (Type _) = False
1240 interestingDict (App fn (Type _)) = interestingDict fn
1241 interestingDict (Note _ a) = interestingDict a
1242 interestingDict (Cast e _) = interestingDict e
1243 interestingDict _ = True
1247 plusUDs :: UsageDetails -> UsageDetails -> UsageDetails
1248 plusUDs (MkUD {dict_binds = db1, calls = calls1, ud_fvs = fvs1})
1249 (MkUD {dict_binds = db2, calls = calls2, ud_fvs = fvs2})
1250 = MkUD {dict_binds = d, calls = c, ud_fvs = fvs1 `unionVarSet` fvs2}
1252 d = db1 `unionBags` db2
1253 c = calls1 `unionCalls` calls2
1255 plusUDList :: [UsageDetails] -> UsageDetails
1256 plusUDList = foldr plusUDs emptyUDs
1258 -- zapCalls deletes calls to ids from uds
1259 zapCalls :: [Id] -> CallDetails -> CallDetails
1260 zapCalls ids calls = delListFromFM calls ids
1262 mkDB :: CoreBind -> DictBind
1263 mkDB bind = (bind, bind_fvs bind)
1265 bind_fvs :: CoreBind -> VarSet
1266 bind_fvs (NonRec bndr rhs) = pair_fvs (bndr,rhs)
1267 bind_fvs (Rec prs) = foldl delVarSet rhs_fvs bndrs
1270 rhs_fvs = unionVarSets (map pair_fvs prs)
1272 pair_fvs :: (Id, CoreExpr) -> VarSet
1273 pair_fvs (bndr, rhs) = exprFreeVars rhs `unionVarSet` idFreeVars bndr
1274 -- Don't forget variables mentioned in the
1275 -- rules of the bndr. C.f. OccAnal.addRuleUsage
1276 -- Also tyvars mentioned in its type; they may not appear in the RHS
1280 addDictBind :: (Id,CoreExpr) -> UsageDetails -> UsageDetails
1281 addDictBind (dict,rhs) uds
1282 = uds { dict_binds = db `consBag` dict_binds uds
1283 , ud_fvs = ud_fvs uds `unionVarSet` fvs }
1285 db@(_, fvs) = mkDB (NonRec dict rhs)
1287 dumpAllDictBinds :: UsageDetails -> [CoreBind] -> [CoreBind]
1288 dumpAllDictBinds (MkUD {dict_binds = dbs}) binds
1289 = foldrBag add binds dbs
1291 add (bind,_) binds = bind : binds
1293 dumpUDs :: [CoreBndr]
1294 -> UsageDetails -> CoreExpr
1295 -> (UsageDetails, CoreExpr)
1296 dumpUDs bndrs (MkUD { dict_binds = orig_dbs
1297 , calls = orig_calls
1298 , ud_fvs = fvs}) body
1299 = (new_uds, foldrBag add_let body dump_dbs)
1300 -- This may delete fewer variables
1301 -- than in priciple possible
1304 MkUD { dict_binds = free_dbs
1305 , calls = free_calls
1306 , ud_fvs = fvs `minusVarSet` bndr_set}
1308 bndr_set = mkVarSet bndrs
1309 add_let (bind,_) body = Let bind body
1311 (free_dbs, dump_dbs, dump_set)
1312 = foldlBag dump_db (emptyBag, emptyBag, bndr_set) orig_dbs
1313 -- Important that it's foldl not foldr;
1314 -- we're accumulating the set of dumped ids in dump_set
1316 free_calls = filterCalls dump_set orig_calls
1318 dump_db (free_dbs, dump_dbs, dump_idset) db@(bind, fvs)
1319 | dump_idset `intersectsVarSet` fvs -- Dump it
1320 = (free_dbs, dump_dbs `snocBag` db,
1321 extendVarSetList dump_idset (bindersOf bind))
1323 | otherwise -- Don't dump it
1324 = (free_dbs `snocBag` db, dump_dbs, dump_idset)
1326 filterCalls :: VarSet -> CallDetails -> CallDetails
1327 -- Remove any calls that mention the variables
1328 filterCalls bs calls
1329 = mapFM (\_ cs -> filter_calls cs) $
1330 filterFM (\k _ -> not (k `elemVarSet` bs)) calls
1332 filter_calls :: CallInfo -> CallInfo
1333 filter_calls = filterFM (\_ (_, fvs) -> not (fvs `intersectsVarSet` bs))
1337 %************************************************************************
1339 \subsubsection{Boring helper functions}
1341 %************************************************************************
1344 type SpecM a = UniqSM a
1346 initSM :: UniqSupply -> SpecM a -> a
1349 mapAndCombineSM :: (a -> SpecM (b, UsageDetails)) -> [a] -> SpecM ([b], UsageDetails)
1350 mapAndCombineSM _ [] = return ([], emptyUDs)
1351 mapAndCombineSM f (x:xs) = do (y, uds1) <- f x
1352 (ys, uds2) <- mapAndCombineSM f xs
1353 return (y:ys, uds1 `plusUDs` uds2)
1355 cloneBindSM :: Subst -> CoreBind -> SpecM (Subst, Subst, CoreBind)
1356 -- Clone the binders of the bind; return new bind with the cloned binders
1357 -- Return the substitution to use for RHSs, and the one to use for the body
1358 cloneBindSM subst (NonRec bndr rhs) = do
1359 us <- getUniqueSupplyM
1360 let (subst', bndr') = cloneIdBndr subst us bndr
1361 return (subst, subst', NonRec bndr' rhs)
1363 cloneBindSM subst (Rec pairs) = do
1364 us <- getUniqueSupplyM
1365 let (subst', bndrs') = cloneRecIdBndrs subst us (map fst pairs)
1366 return (subst', subst', Rec (bndrs' `zip` map snd pairs))
1368 cloneDictBndrs :: Subst -> [CoreBndr] -> SpecM (Subst, [CoreBndr])
1369 cloneDictBndrs subst bndrs
1370 = do { us <- getUniqueSupplyM
1371 ; return (cloneIdBndrs subst us bndrs) }
1373 newSpecIdSM :: Id -> Type -> SpecM Id
1374 -- Give the new Id a similar occurrence name to the old one
1375 newSpecIdSM old_id new_ty
1376 = do { uniq <- getUniqueM
1378 name = idName old_id
1379 new_occ = mkSpecOcc (nameOccName name)
1380 new_id = mkUserLocal new_occ uniq new_ty (getSrcSpan name)
1385 Old (but interesting) stuff about unboxed bindings
1386 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1388 What should we do when a value is specialised to a *strict* unboxed value?
1390 map_*_* f (x:xs) = let h = f x
1394 Could convert let to case:
1396 map_*_Int# f (x:xs) = case f x of h# ->
1400 This may be undesirable since it forces evaluation here, but the value
1401 may not be used in all branches of the body. In the general case this
1402 transformation is impossible since the mutual recursion in a letrec
1403 cannot be expressed as a case.
1405 There is also a problem with top-level unboxed values, since our
1406 implementation cannot handle unboxed values at the top level.
1408 Solution: Lift the binding of the unboxed value and extract it when it
1411 map_*_Int# f (x:xs) = let h = case (f x) of h# -> _Lift h#
1416 Now give it to the simplifier and the _Lifting will be optimised away.
1418 The benfit is that we have given the specialised "unboxed" values a
1419 very simplep lifted semantics and then leave it up to the simplifier to
1420 optimise it --- knowing that the overheads will be removed in nearly
1423 In particular, the value will only be evaluted in the branches of the
1424 program which use it, rather than being forced at the point where the
1425 value is bound. For example:
1427 filtermap_*_* p f (x:xs)
1434 filtermap_*_Int# p f (x:xs)
1435 = let h = case (f x) of h# -> _Lift h#
1438 True -> case h of _Lift h#
1442 The binding for h can still be inlined in the one branch and the
1443 _Lifting eliminated.
1446 Question: When won't the _Lifting be eliminated?
1448 Answer: When they at the top-level (where it is necessary) or when
1449 inlining would duplicate work (or possibly code depending on
1450 options). However, the _Lifting will still be eliminated if the
1451 strictness analyser deems the lifted binding strict.