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,
18 idInlineActivation, setInlineActivation, setIdUnfolding,
19 isLocalId, 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 )
30 import SimplUtils ( interestingArg )
36 import CoreUtils ( exprIsTrivial, applyTypeToArgs, mkPiTypes )
37 import CoreFVs ( exprFreeVars, exprsFreeVars, idFreeVars )
38 import UniqSupply ( UniqSupply,
43 import MkId ( voidArgId, realWorldPrimId )
45 import Maybes ( catMaybes, isJust )
53 %************************************************************************
55 \subsection[notes-Specialise]{Implementation notes [SLPJ, Aug 18 1993]}
57 %************************************************************************
59 These notes describe how we implement specialisation to eliminate
62 The specialisation pass works on Core
63 syntax, complete with all the explicit dictionary application,
64 abstraction and construction as added by the type checker. The
65 existing type checker remains largely as it is.
67 One important thought: the {\em types} passed to an overloaded
68 function, and the {\em dictionaries} passed are mutually redundant.
69 If the same function is applied to the same type(s) then it is sure to
70 be applied to the same dictionary(s)---or rather to the same {\em
71 values}. (The arguments might look different but they will evaluate
74 Second important thought: we know that we can make progress by
75 treating dictionary arguments as static and worth specialising on. So
76 we can do without binding-time analysis, and instead specialise on
77 dictionary arguments and no others.
86 and suppose f is overloaded.
88 STEP 1: CALL-INSTANCE COLLECTION
90 We traverse <body>, accumulating all applications of f to types and
93 (Might there be partial applications, to just some of its types and
94 dictionaries? In principle yes, but in practice the type checker only
95 builds applications of f to all its types and dictionaries, so partial
96 applications could only arise as a result of transformation, and even
97 then I think it's unlikely. In any case, we simply don't accumulate such
98 partial applications.)
103 So now we have a collection of calls to f:
107 Notice that f may take several type arguments. To avoid ambiguity, we
108 say that f is called at type t1/t2 and t3/t4.
110 We take equivalence classes using equality of the *types* (ignoring
111 the dictionary args, which as mentioned previously are redundant).
113 STEP 3: SPECIALISATION
115 For each equivalence class, choose a representative (f t1 t2 d1 d2),
116 and create a local instance of f, defined thus:
118 f@t1/t2 = <f_rhs> t1 t2 d1 d2
120 f_rhs presumably has some big lambdas and dictionary lambdas, so lots
121 of simplification will now result. However we don't actually *do* that
122 simplification. Rather, we leave it for the simplifier to do. If we
123 *did* do it, though, we'd get more call instances from the specialised
124 RHS. We can work out what they are by instantiating the call-instance
125 set from f's RHS with the types t1, t2.
127 Add this new id to f's IdInfo, to record that f has a specialised version.
129 Before doing any of this, check that f's IdInfo doesn't already
130 tell us about an existing instance of f at the required type/s.
131 (This might happen if specialisation was applied more than once, or
132 it might arise from user SPECIALIZE pragmas.)
136 Wait a minute! What if f is recursive? Then we can't just plug in
137 its right-hand side, can we?
139 But it's ok. The type checker *always* creates non-recursive definitions
140 for overloaded recursive functions. For example:
142 f x = f (x+x) -- Yes I know its silly
146 f a (d::Num a) = let p = +.sel a d
148 letrec fl (y::a) = fl (p y y)
152 We still have recusion for non-overloaded functions which we
153 speciailise, but the recursive call should get specialised to the
154 same recursive version.
160 All this is crystal clear when the function is applied to *constant
161 types*; that is, types which have no type variables inside. But what if
162 it is applied to non-constant types? Suppose we find a call of f at type
163 t1/t2. There are two possibilities:
165 (a) The free type variables of t1, t2 are in scope at the definition point
166 of f. In this case there's no problem, we proceed just as before. A common
167 example is as follows. Here's the Haskell:
172 After typechecking we have
174 g a (d::Num a) (y::a) = let f b (d'::Num b) (x::b) = +.sel b d' x x
175 in +.sel a d (f a d y) (f a d y)
177 Notice that the call to f is at type type "a"; a non-constant type.
178 Both calls to f are at the same type, so we can specialise to give:
180 g a (d::Num a) (y::a) = let f@a (x::a) = +.sel a d x x
181 in +.sel a d (f@a y) (f@a y)
184 (b) The other case is when the type variables in the instance types
185 are *not* in scope at the definition point of f. The example we are
186 working with above is a good case. There are two instances of (+.sel a d),
187 but "a" is not in scope at the definition of +.sel. Can we do anything?
188 Yes, we can "common them up", a sort of limited common sub-expression deal.
191 g a (d::Num a) (y::a) = let +.sel@a = +.sel a d
192 f@a (x::a) = +.sel@a x x
193 in +.sel@a (f@a y) (f@a y)
195 This can save work, and can't be spotted by the type checker, because
196 the two instances of +.sel weren't originally at the same type.
200 * There are quite a few variations here. For example, the defn of
201 +.sel could be floated ouside the \y, to attempt to gain laziness.
202 It certainly mustn't be floated outside the \d because the d has to
205 * We don't want to inline f_rhs in this case, because
206 that will duplicate code. Just commoning up the call is the point.
208 * Nothing gets added to +.sel's IdInfo.
210 * Don't bother unless the equivalence class has more than one item!
212 Not clear whether this is all worth it. It is of course OK to
213 simply discard call-instances when passing a big lambda.
215 Polymorphism 2 -- Overloading
217 Consider a function whose most general type is
219 f :: forall a b. Ord a => [a] -> b -> b
221 There is really no point in making a version of g at Int/Int and another
222 at Int/Bool, because it's only instancing the type variable "a" which
223 buys us any efficiency. Since g is completely polymorphic in b there
224 ain't much point in making separate versions of g for the different
227 That suggests that we should identify which of g's type variables
228 are constrained (like "a") and which are unconstrained (like "b").
229 Then when taking equivalence classes in STEP 2, we ignore the type args
230 corresponding to unconstrained type variable. In STEP 3 we make
231 polymorphic versions. Thus:
233 f@t1/ = /\b -> <f_rhs> t1 b d1 d2
242 f a (d::Num a) = let g = ...
244 ...(let d1::Ord a = Num.Ord.sel a d in g a d1)...
246 Here, g is only called at one type, but the dictionary isn't in scope at the
247 definition point for g. Usually the type checker would build a
248 definition for d1 which enclosed g, but the transformation system
249 might have moved d1's defn inward. Solution: float dictionary bindings
250 outwards along with call instances.
254 f x = let g p q = p==q
260 Before specialisation, leaving out type abstractions we have
262 f df x = let g :: Eq a => a -> a -> Bool
264 h :: Num a => a -> a -> (a, Bool)
265 h dh r s = let deq = eqFromNum dh
266 in (+ dh r s, g deq r s)
270 After specialising h we get a specialised version of h, like this:
272 h' r s = let deq = eqFromNum df
273 in (+ df r s, g deq r s)
275 But we can't naively make an instance for g from this, because deq is not in scope
276 at the defn of g. Instead, we have to float out the (new) defn of deq
277 to widen its scope. Notice that this floating can't be done in advance -- it only
278 shows up when specialisation is done.
280 User SPECIALIZE pragmas
281 ~~~~~~~~~~~~~~~~~~~~~~~
282 Specialisation pragmas can be digested by the type checker, and implemented
283 by adding extra definitions along with that of f, in the same way as before
285 f@t1/t2 = <f_rhs> t1 t2 d1 d2
287 Indeed the pragmas *have* to be dealt with by the type checker, because
288 only it knows how to build the dictionaries d1 and d2! For example
290 g :: Ord a => [a] -> [a]
291 {-# SPECIALIZE f :: [Tree Int] -> [Tree Int] #-}
293 Here, the specialised version of g is an application of g's rhs to the
294 Ord dictionary for (Tree Int), which only the type checker can conjure
295 up. There might not even *be* one, if (Tree Int) is not an instance of
296 Ord! (All the other specialision has suitable dictionaries to hand
299 Problem. The type checker doesn't have to hand a convenient <f_rhs>, because
300 it is buried in a complex (as-yet-un-desugared) binding group.
303 f@t1/t2 = f* t1 t2 d1 d2
305 where f* is the Id f with an IdInfo which says "inline me regardless!".
306 Indeed all the specialisation could be done in this way.
307 That in turn means that the simplifier has to be prepared to inline absolutely
308 any in-scope let-bound thing.
311 Again, the pragma should permit polymorphism in unconstrained variables:
313 h :: Ord a => [a] -> b -> b
314 {-# SPECIALIZE h :: [Int] -> b -> b #-}
316 We *insist* that all overloaded type variables are specialised to ground types,
317 (and hence there can be no context inside a SPECIALIZE pragma).
318 We *permit* unconstrained type variables to be specialised to
320 - or left as a polymorphic type variable
321 but nothing in between. So
323 {-# SPECIALIZE h :: [Int] -> [c] -> [c] #-}
325 is *illegal*. (It can be handled, but it adds complication, and gains the
329 SPECIALISING INSTANCE DECLARATIONS
330 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
333 instance Foo a => Foo [a] where
335 {-# SPECIALIZE instance Foo [Int] #-}
337 The original instance decl creates a dictionary-function
340 dfun.Foo.List :: forall a. Foo a -> Foo [a]
342 The SPECIALIZE pragma just makes a specialised copy, just as for
343 ordinary function definitions:
345 dfun.Foo.List@Int :: Foo [Int]
346 dfun.Foo.List@Int = dfun.Foo.List Int dFooInt
348 The information about what instance of the dfun exist gets added to
349 the dfun's IdInfo in the same way as a user-defined function too.
352 Automatic instance decl specialisation?
353 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
354 Can instance decls be specialised automatically? It's tricky.
355 We could collect call-instance information for each dfun, but
356 then when we specialised their bodies we'd get new call-instances
357 for ordinary functions; and when we specialised their bodies, we might get
358 new call-instances of the dfuns, and so on. This all arises because of
359 the unrestricted mutual recursion between instance decls and value decls.
361 Still, there's no actual problem; it just means that we may not do all
362 the specialisation we could theoretically do.
364 Furthermore, instance decls are usually exported and used non-locally,
365 so we'll want to compile enough to get those specialisations done.
367 Lastly, there's no such thing as a local instance decl, so we can
368 survive solely by spitting out *usage* information, and then reading that
369 back in as a pragma when next compiling the file. So for now,
370 we only specialise instance decls in response to pragmas.
373 SPITTING OUT USAGE INFORMATION
374 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
376 To spit out usage information we need to traverse the code collecting
377 call-instance information for all imported (non-prelude?) functions
378 and data types. Then we equivalence-class it and spit it out.
380 This is done at the top-level when all the call instances which escape
381 must be for imported functions and data types.
383 *** Not currently done ***
386 Partial specialisation by pragmas
387 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
388 What about partial specialisation:
390 k :: (Ord a, Eq b) => [a] -> b -> b -> [a]
391 {-# SPECIALIZE k :: Eq b => [Int] -> b -> b -> [a] #-}
395 {-# SPECIALIZE k :: Eq b => [Int] -> [b] -> [b] -> [a] #-}
397 Seems quite reasonable. Similar things could be done with instance decls:
399 instance (Foo a, Foo b) => Foo (a,b) where
401 {-# SPECIALIZE instance Foo a => Foo (a,Int) #-}
402 {-# SPECIALIZE instance Foo b => Foo (Int,b) #-}
404 Ho hum. Things are complex enough without this. I pass.
407 Requirements for the simplifer
408 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
409 The simplifier has to be able to take advantage of the specialisation.
411 * When the simplifier finds an application of a polymorphic f, it looks in
412 f's IdInfo in case there is a suitable instance to call instead. This converts
414 f t1 t2 d1 d2 ===> f_t1_t2
416 Note that the dictionaries get eaten up too!
418 * Dictionary selection operations on constant dictionaries must be
421 +.sel Int d ===> +Int
423 The obvious way to do this is in the same way as other specialised
424 calls: +.sel has inside it some IdInfo which tells that if it's applied
425 to the type Int then it should eat a dictionary and transform to +Int.
427 In short, dictionary selectors need IdInfo inside them for constant
430 * Exactly the same applies if a superclass dictionary is being
433 Eq.sel Int d ===> dEqInt
435 * Something similar applies to dictionary construction too. Suppose
436 dfun.Eq.List is the function taking a dictionary for (Eq a) to
437 one for (Eq [a]). Then we want
439 dfun.Eq.List Int d ===> dEq.List_Int
441 Where does the Eq [Int] dictionary come from? It is built in
442 response to a SPECIALIZE pragma on the Eq [a] instance decl.
444 In short, dfun Ids need IdInfo with a specialisation for each
445 constant instance of their instance declaration.
447 All this uses a single mechanism: the SpecEnv inside an Id
450 What does the specialisation IdInfo look like?
451 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
453 The SpecEnv of an Id maps a list of types (the template) to an expression
457 For example, if f has this SpecInfo:
459 [Int, a] -> \d:Ord Int. f' a
461 it means that we can replace the call
463 f Int t ===> (\d. f' t)
465 This chucks one dictionary away and proceeds with the
466 specialised version of f, namely f'.
469 What can't be done this way?
470 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
471 There is no way, post-typechecker, to get a dictionary for (say)
472 Eq a from a dictionary for Eq [a]. So if we find
476 we can't transform to
481 eqList :: (a->a->Bool) -> [a] -> [a] -> Bool
483 Of course, we currently have no way to automatically derive
484 eqList, nor to connect it to the Eq [a] instance decl, but you
485 can imagine that it might somehow be possible. Taking advantage
486 of this is permanently ruled out.
488 Still, this is no great hardship, because we intend to eliminate
489 overloading altogether anyway!
491 A note about non-tyvar dictionaries
492 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
493 Some Ids have types like
495 forall a,b,c. Eq a -> Ord [a] -> tau
497 This seems curious at first, because we usually only have dictionary
498 args whose types are of the form (C a) where a is a type variable.
499 But this doesn't hold for the functions arising from instance decls,
500 which sometimes get arguements with types of form (C (T a)) for some
503 Should we specialise wrt this compound-type dictionary? We used to say
505 "This is a heuristic judgement, as indeed is the fact that we
506 specialise wrt only dictionaries. We choose *not* to specialise
507 wrt compound dictionaries because at the moment the only place
508 they show up is in instance decls, where they are simply plugged
509 into a returned dictionary. So nothing is gained by specialising
512 But it is simpler and more uniform to specialise wrt these dicts too;
513 and in future GHC is likely to support full fledged type signatures
515 f :: Eq [(a,b)] => ...
518 %************************************************************************
520 \subsubsection{The new specialiser}
522 %************************************************************************
524 Our basic game plan is this. For let(rec) bound function
525 f :: (C a, D c) => (a,b,c,d) -> Bool
527 * Find any specialised calls of f, (f ts ds), where
528 ts are the type arguments t1 .. t4, and
529 ds are the dictionary arguments d1 .. d2.
531 * Add a new definition for f1 (say):
533 f1 = /\ b d -> (..body of f..) t1 b t3 d d1 d2
535 Note that we abstract over the unconstrained type arguments.
539 [t1,b,t3,d] |-> \d1 d2 -> f1 b d
541 to the specialisations of f. This will be used by the
542 simplifier to replace calls
543 (f t1 t2 t3 t4) da db
545 (\d1 d1 -> f1 t2 t4) da db
547 All the stuff about how many dictionaries to discard, and what types
548 to apply the specialised function to, are handled by the fact that the
549 SpecEnv contains a template for the result of the specialisation.
551 We don't build *partial* specialisations for f. For example:
553 f :: Eq a => a -> a -> Bool
554 {-# SPECIALISE f :: (Eq b, Eq c) => (b,c) -> (b,c) -> Bool #-}
556 Here, little is gained by making a specialised copy of f.
557 There's a distinct danger that the specialised version would
558 first build a dictionary for (Eq b, Eq c), and then select the (==)
559 method from it! Even if it didn't, not a great deal is saved.
561 We do, however, generate polymorphic, but not overloaded, specialisations:
563 f :: Eq a => [a] -> b -> b -> b
564 {#- SPECIALISE f :: [Int] -> b -> b -> b #-}
566 Hence, the invariant is this:
568 *** no specialised version is overloaded ***
571 %************************************************************************
573 \subsubsection{The exported function}
575 %************************************************************************
578 specProgram :: UniqSupply -> [CoreBind] -> [CoreBind]
579 specProgram us binds = initSM us (do (binds', uds') <- go binds
580 return (dumpAllDictBinds uds' binds'))
582 -- We need to start with a Subst that knows all the things
583 -- that are in scope, so that the substitution engine doesn't
584 -- accidentally re-use a unique that's already in use
585 -- Easiest thing is to do it all at once, as if all the top-level
586 -- decls were mutually recursive
587 top_subst = mkEmptySubst (mkInScopeSet (mkVarSet (bindersOfBinds binds)))
589 go [] = return ([], emptyUDs)
590 go (bind:binds) = do (binds', uds) <- go binds
591 (bind', uds') <- specBind top_subst bind uds
592 return (bind' ++ binds', uds')
595 %************************************************************************
597 \subsubsection{@specExpr@: the main function}
599 %************************************************************************
602 specVar :: Subst -> Id -> CoreExpr
603 specVar subst v = lookupIdSubst subst v
605 specExpr :: Subst -> CoreExpr -> SpecM (CoreExpr, UsageDetails)
606 -- We carry a substitution down:
607 -- a) we must clone any binding that might float outwards,
608 -- to avoid name clashes
609 -- b) we carry a type substitution to use when analysing
610 -- the RHS of specialised bindings (no type-let!)
612 ---------------- First the easy cases --------------------
613 specExpr subst (Type ty) = return (Type (substTy subst ty), emptyUDs)
614 specExpr subst (Var v) = return (specVar subst v, emptyUDs)
615 specExpr _ (Lit lit) = return (Lit lit, emptyUDs)
616 specExpr subst (Cast e co) = do
617 (e', uds) <- specExpr subst e
618 return ((Cast e' (substTy subst co)), uds)
619 specExpr subst (Note note body) = do
620 (body', uds) <- specExpr subst body
621 return (Note (specNote subst note) body', uds)
624 ---------------- Applications might generate a call instance --------------------
625 specExpr subst expr@(App {})
628 go (App fun arg) args = do (arg', uds_arg) <- specExpr subst arg
629 (fun', uds_app) <- go fun (arg':args)
630 return (App fun' arg', uds_arg `plusUDs` uds_app)
632 go (Var f) args = case specVar subst f of
633 Var f' -> return (Var f', mkCallUDs f' args)
634 e' -> return (e', emptyUDs) -- I don't expect this!
635 go other _ = specExpr subst other
637 ---------------- Lambda/case require dumping of usage details --------------------
638 specExpr subst e@(Lam _ _) = do
639 (body', uds) <- specExpr subst' body
640 let (filtered_uds, body'') = dumpUDs bndrs' uds body'
641 return (mkLams bndrs' body'', filtered_uds)
643 (bndrs, body) = collectBinders e
644 (subst', bndrs') = substBndrs subst bndrs
645 -- More efficient to collect a group of binders together all at once
646 -- and we don't want to split a lambda group with dumped bindings
648 specExpr subst (Case scrut case_bndr ty alts) = do
649 (scrut', uds_scrut) <- specExpr subst scrut
650 (alts', uds_alts) <- mapAndCombineSM spec_alt alts
651 return (Case scrut' case_bndr' (substTy subst ty) alts', uds_scrut `plusUDs` uds_alts)
653 (subst_alt, case_bndr') = substBndr subst case_bndr
654 -- No need to clone case binder; it can't float like a let(rec)
656 spec_alt (con, args, rhs) = do
657 (rhs', uds) <- specExpr subst_rhs rhs
658 let (uds', rhs'') = dumpUDs args uds rhs'
659 return ((con, args', rhs''), uds')
661 (subst_rhs, args') = substBndrs subst_alt args
663 ---------------- Finally, let is the interesting case --------------------
664 specExpr subst (Let bind body) = do
666 (rhs_subst, body_subst, bind') <- cloneBindSM subst bind
668 -- Deal with the body
669 (body', body_uds) <- specExpr body_subst body
671 -- Deal with the bindings
672 (binds', uds) <- specBind rhs_subst bind' body_uds
675 return (foldr Let body' binds', uds)
677 -- Must apply the type substitution to coerceions
678 specNote :: Subst -> Note -> Note
679 specNote _ note = note
682 %************************************************************************
684 \subsubsection{Dealing with a binding}
686 %************************************************************************
689 specBind :: Subst -- Use this for RHSs
691 -> UsageDetails -- Info on how the scope of the binding
692 -> SpecM ([CoreBind], -- New bindings
693 UsageDetails) -- And info to pass upstream
695 specBind rhs_subst bind body_uds
696 = do { (bind', bind_uds) <- specBindItself rhs_subst bind (calls body_uds)
697 ; return (finishSpecBind bind' bind_uds body_uds) }
699 finishSpecBind :: CoreBind -> UsageDetails -> UsageDetails -> ([CoreBind], UsageDetails)
701 (MkUD { dict_binds = rhs_dbs, calls = rhs_calls, ud_fvs = rhs_fvs })
702 (MkUD { dict_binds = body_dbs, calls = body_calls, ud_fvs = body_fvs })
703 | not (mkVarSet bndrs `intersectsVarSet` all_fvs)
704 -- Common case 1: the bound variables are not
705 -- mentioned in the dictionary bindings
706 = ([bind], MkUD { dict_binds = body_dbs `unionBags` rhs_dbs
707 -- It's important that the `unionBags` is this way round,
708 -- because body_uds may bind dictionaries that are
709 -- used in the calls passed to specDefn. So the
710 -- dictionary bindings in rhs_uds may mention
711 -- dictionaries bound in body_uds.
713 , ud_fvs = all_fvs })
715 | case bind of { NonRec {} -> True; Rec {} -> False }
716 -- Common case 2: no specialisation happened, and binding
717 -- is non-recursive. But the binding may be
718 -- mentioned in body_dbs, so we should put it first
719 = ([], MkUD { dict_binds = rhs_dbs `unionBags` ((bind, b_fvs) `consBag` body_dbs)
721 , ud_fvs = all_fvs `unionVarSet` b_fvs })
723 | otherwise -- General case: make a huge Rec (sigh)
724 = ([], MkUD { dict_binds = unitBag (Rec all_db_prs, all_db_fvs)
726 , ud_fvs = all_fvs `unionVarSet` b_fvs })
728 all_fvs = rhs_fvs `unionVarSet` body_fvs
729 all_calls = zapCalls bndrs (rhs_calls `unionCalls` body_calls)
731 bndrs = bindersOf bind
732 b_fvs = bind_fvs bind
734 (all_db_prs, all_db_fvs) = add (bind, b_fvs) $
735 foldrBag add ([], emptyVarSet) $
736 rhs_dbs `unionBags` body_dbs
737 add (NonRec b r, b_fvs) (prs, fvs) = ((b,r) : prs, b_fvs `unionVarSet` fvs)
738 add (Rec b_prs, b_fvs) (prs, fvs) = (b_prs ++ prs, b_fvs `unionVarSet` fvs)
740 ---------------------------
741 specBindItself :: Subst -> CoreBind -> CallDetails -> SpecM (CoreBind, UsageDetails)
743 -- specBindItself deals with the RHS, specialising it according
744 -- to the calls found in the body (if any)
745 specBindItself rhs_subst (NonRec fn rhs) call_info
746 = do { (rhs', rhs_uds) <- specExpr rhs_subst rhs -- Do RHS of original fn
747 ; (fn', spec_defns, spec_uds) <- specDefn rhs_subst call_info fn rhs
748 ; if null spec_defns then
749 return (NonRec fn rhs', rhs_uds)
751 return (Rec ((fn',rhs') : spec_defns), rhs_uds `plusUDs` spec_uds) }
752 -- bndr' mentions the spec_defns in its SpecEnv
753 -- Not sure why we couln't just put the spec_defns first
755 specBindItself rhs_subst (Rec pairs) call_info
756 -- Note [Specialising a recursive group]
757 = do { let (bndrs,rhss) = unzip pairs
758 ; (rhss', rhs_uds) <- mapAndCombineSM (specExpr rhs_subst) rhss
759 ; let all_calls = call_info `unionCalls` calls rhs_uds
760 ; (bndrs1, spec_defns1, spec_uds1) <- specDefns rhs_subst all_calls pairs
762 ; if null spec_defns1 then -- Common case: no specialisation
763 return (Rec (bndrs `zip` rhss'), rhs_uds)
764 else do -- Specialisation occurred; do it again
765 { (bndrs2, spec_defns2, spec_uds2) <-
766 -- pprTrace "specB" (ppr bndrs $$ ppr rhs_uds) $
767 specDefns rhs_subst (calls spec_uds1) (bndrs1 `zip` rhss)
769 ; let all_defns = spec_defns1 ++ spec_defns2 ++ zip bndrs2 rhss'
771 ; return (Rec all_defns, rhs_uds `plusUDs` spec_uds1 `plusUDs` spec_uds2) } }
774 ---------------------------
776 -> CallDetails -- Info on how it is used in its scope
777 -> [(Id,CoreExpr)] -- The things being bound and their un-processed RHS
778 -> SpecM ([Id], -- Original Ids with RULES added
779 [(Id,CoreExpr)], -- Extra, specialised bindings
780 UsageDetails) -- Stuff to fling upwards from the specialised versions
782 -- Specialise a list of bindings (the contents of a Rec), but flowing usages
783 -- upwards binding by binding. Example: { f = ...g ...; g = ...f .... }
784 -- Then if the input CallDetails has a specialised call for 'g', whose specialisation
785 -- in turn generates a specialised call for 'f', we catch that in this one sweep.
786 -- But not vice versa (it's a fixpoint problem).
788 specDefns _subst _call_info []
789 = return ([], [], emptyUDs)
790 specDefns subst call_info ((bndr,rhs):pairs)
791 = do { (bndrs', spec_defns, spec_uds) <- specDefns subst call_info pairs
792 ; let all_calls = call_info `unionCalls` calls spec_uds
793 ; (bndr', spec_defns1, spec_uds1) <- specDefn subst all_calls bndr rhs
794 ; return (bndr' : bndrs',
795 spec_defns1 ++ spec_defns,
796 spec_uds1 `plusUDs` spec_uds) }
798 ---------------------------
800 -> CallDetails -- Info on how it is used in its scope
801 -> Id -> CoreExpr -- The thing being bound and its un-processed RHS
802 -> SpecM (Id, -- Original Id with added RULES
803 [(Id,CoreExpr)], -- Extra, specialised bindings
804 UsageDetails) -- Stuff to fling upwards from the specialised versions
806 specDefn subst calls fn rhs
807 -- The first case is the interesting one
808 | rhs_tyvars `lengthIs` n_tyvars -- Rhs of fn's defn has right number of big lambdas
809 && rhs_ids `lengthAtLeast` n_dicts -- and enough dict args
810 && notNull calls_for_me -- And there are some calls to specialise
812 -- && not (certainlyWillInline (idUnfolding fn)) -- And it's not small
813 -- See Note [Inline specialisation] for why we do not
814 -- switch off specialisation for inline functions
816 = do { -- Make a specialised version for each call in calls_for_me
817 stuff <- mapM spec_call calls_for_me
818 ; let (spec_defns, spec_uds, spec_rules) = unzip3 (catMaybes stuff)
819 fn' = addIdSpecialisations fn spec_rules
820 ; return (fn', spec_defns, plusUDList spec_uds) }
822 | otherwise -- No calls or RHS doesn't fit our preconceptions
823 = WARN( notNull calls_for_me, ptext (sLit "Missed specialisation opportunity for") <+> ppr fn )
824 -- Note [Specialisation shape]
825 return (fn, [], emptyUDs)
829 fn_arity = idArity fn
830 (tyvars, theta, _) = tcSplitSigmaTy fn_type
831 n_tyvars = length tyvars
832 n_dicts = length theta
833 inline_act = idInlineActivation fn
835 -- It's important that we "see past" any INLINE pragma
836 -- else we'll fail to specialise an INLINE thing
837 (inline_rhs, rhs_inside) = dropInline rhs
838 (rhs_tyvars, rhs_ids, rhs_body) = collectTyAndValBinders rhs_inside
840 rhs_dict_ids = take n_dicts rhs_ids
841 body = mkLams (drop n_dicts rhs_ids) rhs_body
842 -- Glue back on the non-dict lambdas
844 calls_for_me = case lookupFM calls fn of
846 Just cs -> fmToList cs
848 already_covered :: [CoreExpr] -> Bool
849 already_covered args -- Note [Specialisations already covered]
850 = isJust (lookupRule (const True) (substInScope subst)
851 fn args (idCoreRules fn))
853 mk_ty_args :: [Maybe Type] -> [CoreExpr]
854 mk_ty_args call_ts = zipWithEqual "spec_call" mk_ty_arg rhs_tyvars call_ts
856 mk_ty_arg rhs_tyvar Nothing = Type (mkTyVarTy rhs_tyvar)
857 mk_ty_arg _ (Just ty) = Type ty
859 ----------------------------------------------------------
860 -- Specialise to one particular call pattern
861 spec_call :: (CallKey, ([DictExpr], VarSet)) -- Call instance
862 -> SpecM (Maybe ((Id,CoreExpr), -- Specialised definition
863 UsageDetails, -- Usage details from specialised body
864 CoreRule)) -- Info for the Id's SpecEnv
865 spec_call (CallKey call_ts, (call_ds, _))
866 = ASSERT( call_ts `lengthIs` n_tyvars && call_ds `lengthIs` n_dicts )
868 -- Suppose f's defn is f = /\ a b c -> \ d1 d2 -> rhs
869 -- Supppose the call is for f [Just t1, Nothing, Just t3] [dx1, dx2]
871 -- Construct the new binding
872 -- f1 = SUBST[a->t1,c->t3, d1->d1', d2->d2'] (/\ b -> rhs)
873 -- PLUS the usage-details
874 -- { d1' = dx1; d2' = dx2 }
875 -- where d1', d2' are cloned versions of d1,d2, with the type substitution
876 -- applied. These auxiliary bindings just avoid duplication of dx1, dx2
878 -- Note that the substitution is applied to the whole thing.
879 -- This is convenient, but just slightly fragile. Notably:
880 -- * There had better be no name clashes in a/b/c
882 -- poly_tyvars = [b] in the example above
883 -- spec_tyvars = [a,c]
884 -- ty_args = [t1,b,t3]
885 poly_tyvars = [tv | (tv, Nothing) <- rhs_tyvars `zip` call_ts]
886 spec_tv_binds = [(tv,ty) | (tv, Just ty) <- rhs_tyvars `zip` call_ts]
887 spec_ty_args = map snd spec_tv_binds
888 ty_args = mk_ty_args call_ts
889 rhs_subst = extendTvSubstList subst spec_tv_binds
891 ; (rhs_subst1, inst_dict_ids) <- cloneDictBndrs rhs_subst rhs_dict_ids
892 -- Clone rhs_dicts, including instantiating their types
894 ; let (rhs_subst2, dx_binds) = bindAuxiliaryDicts rhs_subst1 $
895 (my_zipEqual rhs_dict_ids inst_dict_ids call_ds)
896 inst_args = ty_args ++ map Var inst_dict_ids
898 ; if already_covered inst_args then
901 { -- Figure out the type of the specialised function
902 let body_ty = applyTypeToArgs rhs fn_type inst_args
903 (lam_args, app_args) -- Add a dummy argument if body_ty is unlifted
904 | isUnLiftedType body_ty -- C.f. WwLib.mkWorkerArgs
905 = (poly_tyvars ++ [voidArgId], poly_tyvars ++ [realWorldPrimId])
906 | otherwise = (poly_tyvars, poly_tyvars)
907 spec_id_ty = mkPiTypes lam_args body_ty
909 ; spec_f <- newSpecIdSM fn spec_id_ty
910 ; let spec_f_w_arity = setIdArity spec_f (max 0 (fn_arity - n_dicts))
911 -- Adding arity information just propagates it a bit faster
912 -- See Note [Arity decrease] in Simplify
914 ; (spec_rhs, rhs_uds) <- specExpr rhs_subst2 (mkLams lam_args body)
916 -- The rule to put in the function's specialisation is:
917 -- forall b, d1',d2'. f t1 b t3 d1' d2' = f1 b
918 rule_name = mkFastString ("SPEC " ++ showSDoc (ppr fn <+> ppr spec_ty_args))
919 spec_env_rule = mkLocalRule
921 inline_act -- Note [Auto-specialisation and RULES]
923 (poly_tyvars ++ inst_dict_ids)
925 (mkVarApps (Var spec_f_w_arity) app_args)
927 -- Add the { d1' = dx1; d2' = dx2 } usage stuff
928 final_uds = foldr addDictBind rhs_uds dx_binds
930 spec_pr | inline_rhs = (spec_f_w_arity `setInlineActivation` inline_act, Note InlineMe spec_rhs)
931 | otherwise = (spec_f_w_arity, spec_rhs)
933 ; return (Just (spec_pr, final_uds, spec_env_rule)) } }
936 | debugIsOn && not (equalLength xs ys && equalLength ys zs)
937 = pprPanic "my_zipEqual" (vcat [ ppr xs, ppr ys
938 , ppr fn <+> ppr call_ts
939 , ppr (idType fn), ppr theta
940 , ppr n_dicts, ppr rhs_dict_ids
942 | otherwise = zip3 xs ys zs
946 -> [(DictId,DictId,CoreExpr)] -- (orig_dict, inst_dict, dx)
947 -> (Subst, -- Substitute for all orig_dicts
948 [(DictId, CoreExpr)]) -- Auxiliary bindings
949 -- Bind any dictionary arguments to fresh names, to preserve sharing
950 -- Substitution already substitutes orig_dict -> inst_dict
951 bindAuxiliaryDicts subst triples = go subst [] triples
953 go subst binds [] = (subst, binds)
954 go subst binds ((d, dx_id, dx) : pairs)
955 | exprIsTrivial dx = go (extendIdSubst subst d dx) binds pairs
956 -- No auxiliary binding necessary
957 | otherwise = go subst_w_unf ((dx_id,dx) : binds) pairs
959 dx_id1 = dx_id `setIdUnfolding` mkUnfolding False dx
960 subst_w_unf = extendIdSubst subst d (Var dx_id1)
961 -- Important! We're going to substitute dx_id1 for d
962 -- and we want it to look "interesting", else we won't gather *any*
963 -- consequential calls. E.g.
965 -- If we specialise f for a call (f (dfun dNumInt)), we'll get
966 -- a consequent call (g d') with an auxiliary definition
968 -- We want that consequent call to look interesting
971 Note [Specialising a recursive group]
972 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
974 let rec { f x = ...g x'...
975 ; g y = ...f y'.... }
977 Here we specialise 'f' at Char; but that is very likely to lead to
978 a specialisation of 'g' at Char. We must do the latter, else the
979 whole point of specialisation is lost.
981 But we do not want to keep iterating to a fixpoint, because in the
982 presence of polymorphic recursion we might generate an infinite number
985 So we use the following heuristic:
986 * Arrange the rec block in dependency order, so far as possible
987 (the occurrence analyser already does this)
989 * Specialise it much like a sequence of lets
991 * Then go through the block a second time, feeding call-info from
992 the RHSs back in the bottom, as it were
994 In effect, the ordering maxmimises the effectiveness of each sweep,
995 and we do just two sweeps. This should catch almost every case of
996 monomorphic recursion -- the exception could be a very knotted-up
997 recursion with multiple cycles tied up together.
999 This plan is implemented in the Rec case of specBindItself.
1001 Note [Specialisations already covered]
1002 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1003 We obviously don't want to generate two specialisations for the same
1004 argument pattern. There are two wrinkles
1006 1. We do the already-covered test in specDefn, not when we generate
1007 the CallInfo in mkCallUDs. We used to test in the latter place, but
1008 we now iterate the specialiser somewhat, and the Id at the call site
1009 might therefore not have all the RULES that we can see in specDefn
1011 2. What about two specialisations where the second is an *instance*
1012 of the first? If the more specific one shows up first, we'll generate
1013 specialisations for both. If the *less* specific one shows up first,
1014 we *don't* currently generate a specialisation for the more specific
1015 one. (See the call to lookupRule in already_covered.) Reasons:
1016 (a) lookupRule doesn't say which matches are exact (bad reason)
1017 (b) if the earlier specialisation is user-provided, it's
1018 far from clear that we should auto-specialise further
1020 Note [Auto-specialisation and RULES]
1021 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1023 g :: Num a => a -> a
1026 f :: (Int -> Int) -> Int
1028 {-# RULE f g = 0 #-}
1030 Suppose that auto-specialisation makes a specialised version of
1031 g::Int->Int That version won't appear in the LHS of the RULE for f.
1032 So if the specialisation rule fires too early, the rule for f may
1035 It might be possible to add new rules, to "complete" the rewrite system.
1037 RULE forall d. g Int d = g_spec
1041 But that's a bit complicated. For now we ask the programmer's help,
1042 by *copying the INLINE activation pragma* to the auto-specialised rule.
1043 So if g says {-# NOINLINE[2] g #-}, then the auto-spec rule will also
1044 not be active until phase 2.
1047 Note [Specialisation shape]
1048 ~~~~~~~~~~~~~~~~~~~~~~~~~~~
1049 We only specialise a function if it has visible top-level lambdas
1050 corresponding to its overloading. E.g. if
1051 f :: forall a. Eq a => ....
1052 then its body must look like
1055 Reason: when specialising the body for a call (f ty dexp), we want to
1056 substitute dexp for d, and pick up specialised calls in the body of f.
1058 This doesn't always work. One example I came across was this:
1059 newtype Gen a = MkGen{ unGen :: Int -> a }
1061 choose :: Eq a => a -> Gen a
1062 choose n = MkGen (\r -> n)
1064 oneof = choose (1::Int)
1066 It's a silly exapmle, but we get
1067 choose = /\a. g `cast` co
1068 where choose doesn't have any dict arguments. Thus far I have not
1069 tried to fix this (wait till there's a real example).
1072 Note [Inline specialisations]
1073 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1074 We transfer to the specialised function any INLINE stuff from the
1075 original. This means (a) the Activation in the IdInfo, and (b) any
1076 InlineMe on the RHS. We do not, however, transfer the RuleMatchInfo
1077 since we do not expect the specialisation to occur in rewrite rules.
1079 This is a change (Jun06). Previously the idea is that the point of
1080 inlining was precisely to specialise the function at its call site,
1081 and that's not so important for the specialised copies. But
1082 *pragma-directed* specialisation now takes place in the
1083 typechecker/desugarer, with manually specified INLINEs. The
1084 specialiation here is automatic. It'd be very odd if a function
1085 marked INLINE was specialised (because of some local use), and then
1086 forever after (including importing modules) the specialised version
1087 wasn't INLINEd. After all, the programmer said INLINE!
1089 You might wonder why we don't just not specialise INLINE functions.
1090 It's because even INLINE functions are sometimes not inlined, when
1091 they aren't applied to interesting arguments. But perhaps the type
1092 arguments alone are enough to specialise (even though the args are too
1093 boring to trigger inlining), and it's certainly better to call the
1094 specialised version.
1096 A case in point is dictionary functions, which are current marked
1097 INLINE, but which are worth specialising.
1100 dropInline :: CoreExpr -> (Bool, CoreExpr)
1101 dropInline (Note InlineMe rhs) = (True, rhs)
1102 dropInline rhs = (False, rhs)
1105 %************************************************************************
1107 \subsubsection{UsageDetails and suchlike}
1109 %************************************************************************
1114 dict_binds :: !(Bag DictBind),
1115 -- Floated dictionary bindings
1116 -- The order is important;
1117 -- in ds1 `union` ds2, bindings in ds2 can depend on those in ds1
1118 -- (Remember, Bags preserve order in GHC.)
1120 calls :: !CallDetails,
1122 ud_fvs :: !VarSet -- A superset of the variables mentioned in
1123 -- either dict_binds or calls
1126 instance Outputable UsageDetails where
1127 ppr (MkUD { dict_binds = dbs, calls = calls, ud_fvs = fvs })
1128 = ptext (sLit "MkUD") <+> braces (sep (punctuate comma
1129 [ptext (sLit "binds") <+> equals <+> ppr dbs,
1130 ptext (sLit "calls") <+> equals <+> ppr calls,
1131 ptext (sLit "fvs") <+> equals <+> ppr fvs]))
1133 type DictBind = (CoreBind, VarSet)
1134 -- The set is the free vars of the binding
1135 -- both tyvars and dicts
1137 type DictExpr = CoreExpr
1139 emptyUDs :: UsageDetails
1140 emptyUDs = MkUD { dict_binds = emptyBag, calls = emptyFM, ud_fvs = emptyVarSet }
1142 ------------------------------------------------------------
1143 type CallDetails = FiniteMap Id CallInfo
1144 newtype CallKey = CallKey [Maybe Type] -- Nothing => unconstrained type argument
1146 -- CallInfo uses a FiniteMap, thereby ensuring that
1147 -- we record only one call instance for any key
1149 -- The list of types and dictionaries is guaranteed to
1150 -- match the type of f
1151 type CallInfo = FiniteMap CallKey ([DictExpr], VarSet)
1152 -- Range is dict args and the vars of the whole
1153 -- call (including tyvars)
1154 -- [*not* include the main id itself, of course]
1156 instance Outputable CallKey where
1157 ppr (CallKey ts) = ppr ts
1159 -- Type isn't an instance of Ord, so that we can control which
1160 -- instance we use. That's tiresome here. Oh well
1161 instance Eq CallKey where
1162 k1 == k2 = case k1 `compare` k2 of { EQ -> True; _ -> False }
1164 instance Ord CallKey where
1165 compare (CallKey k1) (CallKey k2) = cmpList cmp k1 k2
1167 cmp Nothing Nothing = EQ
1168 cmp Nothing (Just _) = LT
1169 cmp (Just _) Nothing = GT
1170 cmp (Just t1) (Just t2) = tcCmpType t1 t2
1172 unionCalls :: CallDetails -> CallDetails -> CallDetails
1173 unionCalls c1 c2 = plusFM_C plusFM c1 c2
1175 singleCall :: Id -> [Maybe Type] -> [DictExpr] -> UsageDetails
1176 singleCall id tys dicts
1177 = MkUD {dict_binds = emptyBag,
1178 calls = unitFM id (unitFM (CallKey tys) (dicts, call_fvs)),
1181 call_fvs = exprsFreeVars dicts `unionVarSet` tys_fvs
1182 tys_fvs = tyVarsOfTypes (catMaybes tys)
1183 -- The type args (tys) are guaranteed to be part of the dictionary
1184 -- types, because they are just the constrained types,
1185 -- and the dictionary is therefore sure to be bound
1186 -- inside the binding for any type variables free in the type;
1187 -- hence it's safe to neglect tyvars free in tys when making
1188 -- the free-var set for this call
1189 -- BUT I don't trust this reasoning; play safe and include tys_fvs
1191 -- We don't include the 'id' itself.
1193 mkCallUDs :: Id -> [CoreExpr] -> UsageDetails
1195 | not (isLocalId f) -- Imported from elsewhere
1196 || null theta -- Not overloaded
1197 || not (all isClassPred theta)
1198 -- Only specialise if all overloading is on class params.
1199 -- In ptic, with implicit params, the type args
1200 -- *don't* say what the value of the implicit param is!
1201 || not (spec_tys `lengthIs` n_tyvars)
1202 || not ( dicts `lengthIs` n_dicts)
1203 || not (any interestingArg dicts) -- Note [Interesting dictionary arguments]
1204 -- See also Note [Specialisations already covered]
1205 = -- pprTrace "mkCallUDs: discarding" (vcat [ppr f, ppr args, ppr n_tyvars, ppr n_dicts, ppr (map interestingArg dicts)])
1206 emptyUDs -- Not overloaded, or no specialisation wanted
1209 = -- pprTrace "mkCallUDs: keeping" (vcat [ppr f, ppr args, ppr n_tyvars, ppr n_dicts, ppr (map interestingArg dicts)])
1210 singleCall f spec_tys dicts
1212 (tyvars, theta, _) = tcSplitSigmaTy (idType f)
1213 constrained_tyvars = tyVarsOfTheta theta
1214 n_tyvars = length tyvars
1215 n_dicts = length theta
1217 spec_tys = [mk_spec_ty tv ty | (tv, Type ty) <- tyvars `zip` args]
1218 dicts = [dict_expr | (_, dict_expr) <- theta `zip` (drop n_tyvars args)]
1221 | tyvar `elemVarSet` constrained_tyvars = Just ty
1222 | otherwise = Nothing
1225 Note [Interesting dictionary arguments]
1226 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1228 \a.\d:Eq a. let f = ... in ...(f d)...
1229 There really is not much point in specialising f wrt the dictionary d,
1230 because the code for the specialised f is not improved at all, because
1231 d is lambda-bound. We simply get junk specialisations.
1233 We re-use the function SimplUtils.interestingArg function to determine
1234 what sort of dictionary arguments have *some* information in them.
1238 plusUDs :: UsageDetails -> UsageDetails -> UsageDetails
1239 plusUDs (MkUD {dict_binds = db1, calls = calls1, ud_fvs = fvs1})
1240 (MkUD {dict_binds = db2, calls = calls2, ud_fvs = fvs2})
1241 = MkUD {dict_binds = d, calls = c, ud_fvs = fvs1 `unionVarSet` fvs2}
1243 d = db1 `unionBags` db2
1244 c = calls1 `unionCalls` calls2
1246 plusUDList :: [UsageDetails] -> UsageDetails
1247 plusUDList = foldr plusUDs emptyUDs
1249 -- zapCalls deletes calls to ids from uds
1250 zapCalls :: [Id] -> CallDetails -> CallDetails
1251 zapCalls ids calls = delListFromFM calls ids
1253 mkDB :: CoreBind -> DictBind
1254 mkDB bind = (bind, bind_fvs bind)
1256 bind_fvs :: CoreBind -> VarSet
1257 bind_fvs (NonRec bndr rhs) = pair_fvs (bndr,rhs)
1258 bind_fvs (Rec prs) = foldl delVarSet rhs_fvs bndrs
1261 rhs_fvs = unionVarSets (map pair_fvs prs)
1263 pair_fvs :: (Id, CoreExpr) -> VarSet
1264 pair_fvs (bndr, rhs) = exprFreeVars rhs `unionVarSet` idFreeVars bndr
1265 -- Don't forget variables mentioned in the
1266 -- rules of the bndr. C.f. OccAnal.addRuleUsage
1267 -- Also tyvars mentioned in its type; they may not appear in the RHS
1271 addDictBind :: (Id,CoreExpr) -> UsageDetails -> UsageDetails
1272 addDictBind (dict,rhs) uds
1273 = uds { dict_binds = db `consBag` dict_binds uds
1274 , ud_fvs = ud_fvs uds `unionVarSet` fvs }
1276 db@(_, fvs) = mkDB (NonRec dict rhs)
1278 dumpAllDictBinds :: UsageDetails -> [CoreBind] -> [CoreBind]
1279 dumpAllDictBinds (MkUD {dict_binds = dbs}) binds
1280 = foldrBag add binds dbs
1282 add (bind,_) binds = bind : binds
1284 dumpUDs :: [CoreBndr]
1285 -> UsageDetails -> CoreExpr
1286 -> (UsageDetails, CoreExpr)
1287 dumpUDs bndrs (MkUD { dict_binds = orig_dbs
1288 , calls = orig_calls
1289 , ud_fvs = fvs}) body
1290 = (new_uds, foldrBag add_let body dump_dbs)
1291 -- This may delete fewer variables
1292 -- than in priciple possible
1295 MkUD { dict_binds = free_dbs
1296 , calls = free_calls
1297 , ud_fvs = fvs `minusVarSet` bndr_set}
1299 bndr_set = mkVarSet bndrs
1300 add_let (bind,_) body = Let bind body
1302 (free_dbs, dump_dbs, dump_set)
1303 = foldlBag dump_db (emptyBag, emptyBag, bndr_set) orig_dbs
1304 -- Important that it's foldl not foldr;
1305 -- we're accumulating the set of dumped ids in dump_set
1307 free_calls = filterCalls dump_set orig_calls
1309 dump_db (free_dbs, dump_dbs, dump_idset) db@(bind, fvs)
1310 | dump_idset `intersectsVarSet` fvs -- Dump it
1311 = (free_dbs, dump_dbs `snocBag` db,
1312 extendVarSetList dump_idset (bindersOf bind))
1314 | otherwise -- Don't dump it
1315 = (free_dbs `snocBag` db, dump_dbs, dump_idset)
1317 filterCalls :: VarSet -> CallDetails -> CallDetails
1318 -- Remove any calls that mention the variables
1319 filterCalls bs calls
1320 = mapFM (\_ cs -> filter_calls cs) $
1321 filterFM (\k _ -> not (k `elemVarSet` bs)) calls
1323 filter_calls :: CallInfo -> CallInfo
1324 filter_calls = filterFM (\_ (_, fvs) -> not (fvs `intersectsVarSet` bs))
1328 %************************************************************************
1330 \subsubsection{Boring helper functions}
1332 %************************************************************************
1335 type SpecM a = UniqSM a
1337 initSM :: UniqSupply -> SpecM a -> a
1340 mapAndCombineSM :: (a -> SpecM (b, UsageDetails)) -> [a] -> SpecM ([b], UsageDetails)
1341 mapAndCombineSM _ [] = return ([], emptyUDs)
1342 mapAndCombineSM f (x:xs) = do (y, uds1) <- f x
1343 (ys, uds2) <- mapAndCombineSM f xs
1344 return (y:ys, uds1 `plusUDs` uds2)
1346 cloneBindSM :: Subst -> CoreBind -> SpecM (Subst, Subst, CoreBind)
1347 -- Clone the binders of the bind; return new bind with the cloned binders
1348 -- Return the substitution to use for RHSs, and the one to use for the body
1349 cloneBindSM subst (NonRec bndr rhs) = do
1350 us <- getUniqueSupplyM
1351 let (subst', bndr') = cloneIdBndr subst us bndr
1352 return (subst, subst', NonRec bndr' rhs)
1354 cloneBindSM subst (Rec pairs) = do
1355 us <- getUniqueSupplyM
1356 let (subst', bndrs') = cloneRecIdBndrs subst us (map fst pairs)
1357 return (subst', subst', Rec (bndrs' `zip` map snd pairs))
1359 cloneDictBndrs :: Subst -> [CoreBndr] -> SpecM (Subst, [CoreBndr])
1360 cloneDictBndrs subst bndrs
1361 = do { us <- getUniqueSupplyM
1362 ; return (cloneIdBndrs subst us bndrs) }
1364 newSpecIdSM :: Id -> Type -> SpecM Id
1365 -- Give the new Id a similar occurrence name to the old one
1366 newSpecIdSM old_id new_ty
1367 = do { uniq <- getUniqueM
1369 name = idName old_id
1370 new_occ = mkSpecOcc (nameOccName name)
1371 new_id = mkUserLocal new_occ uniq new_ty (getSrcSpan name)
1376 Old (but interesting) stuff about unboxed bindings
1377 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1379 What should we do when a value is specialised to a *strict* unboxed value?
1381 map_*_* f (x:xs) = let h = f x
1385 Could convert let to case:
1387 map_*_Int# f (x:xs) = case f x of h# ->
1391 This may be undesirable since it forces evaluation here, but the value
1392 may not be used in all branches of the body. In the general case this
1393 transformation is impossible since the mutual recursion in a letrec
1394 cannot be expressed as a case.
1396 There is also a problem with top-level unboxed values, since our
1397 implementation cannot handle unboxed values at the top level.
1399 Solution: Lift the binding of the unboxed value and extract it when it
1402 map_*_Int# f (x:xs) = let h = case (f x) of h# -> _Lift h#
1407 Now give it to the simplifier and the _Lifting will be optimised away.
1409 The benfit is that we have given the specialised "unboxed" values a
1410 very simplep lifted semantics and then leave it up to the simplifier to
1411 optimise it --- knowing that the overheads will be removed in nearly
1414 In particular, the value will only be evaluted in the branches of the
1415 program which use it, rather than being forced at the point where the
1416 value is bound. For example:
1418 filtermap_*_* p f (x:xs)
1425 filtermap_*_Int# p f (x:xs)
1426 = let h = case (f x) of h# -> _Lift h#
1429 True -> case h of _Lift h#
1433 The binding for h can still be inlined in the one branch and the
1434 _Lifting eliminated.
1437 Question: When won't the _Lifting be eliminated?
1439 Answer: When they at the top-level (where it is necessary) or when
1440 inlining would duplicate work (or possibly code depending on
1441 options). However, the _Lifting will still be eliminated if the
1442 strictness analyser deems the lifted binding strict.