2 % (c) The GRASP/AQUA Project, Glasgow University, 1992-1998
4 \section[SpecConstr]{Specialise over constructors}
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
17 #include "HsVersions.h"
22 import CoreUnfold ( couldBeSmallEnoughToInline )
23 import CoreLint ( showPass, endPass )
24 import CoreFVs ( exprsFreeVars )
25 import CoreTidy ( tidyRules )
26 import PprCore ( pprRules )
27 import WwLib ( mkWorkerArgs )
28 import DataCon ( dataConRepArity, dataConUnivTyVars )
30 import Type hiding( substTy )
31 import Id ( Id, idName, idType, isDataConWorkId_maybe, idArity,
32 mkUserLocal, mkSysLocal, idUnfolding, isLocalId )
37 import Rules ( addIdSpecialisations, mkLocalRule, rulesOfBinds )
38 import OccName ( mkSpecOcc )
39 import ErrUtils ( dumpIfSet_dyn )
40 import DynFlags ( DynFlags(..), DynFlag(..) )
41 import StaticFlags ( opt_SpecInlineJoinPoints )
42 import BasicTypes ( Activation(..) )
43 import Maybes ( orElse, catMaybes, isJust, isNothing )
45 import List ( nubBy, partition )
51 import Control.Monad ( zipWithM )
54 -----------------------------------------------------
56 -----------------------------------------------------
61 drop n (x:xs) = drop (n-1) xs
63 After the first time round, we could pass n unboxed. This happens in
64 numerical code too. Here's what it looks like in Core:
66 drop n xs = case xs of
71 _ -> drop (I# (n# -# 1#)) xs
73 Notice that the recursive call has an explicit constructor as argument.
74 Noticing this, we can make a specialised version of drop
76 RULE: drop (I# n#) xs ==> drop' n# xs
78 drop' n# xs = let n = I# n# in ...orig RHS...
80 Now the simplifier will apply the specialisation in the rhs of drop', giving
82 drop' n# xs = case xs of
86 _ -> drop (n# -# 1#) xs
90 We'd also like to catch cases where a parameter is carried along unchanged,
91 but evaluated each time round the loop:
93 f i n = if i>0 || i>n then i else f (i*2) n
95 Here f isn't strict in n, but we'd like to avoid evaluating it each iteration.
96 In Core, by the time we've w/wd (f is strict in i) we get
98 f i# n = case i# ># 0 of
100 True -> case n of n' { I# n# ->
103 True -> f (i# *# 2#) n'
105 At the call to f, we see that the argument, n is know to be (I# n#),
106 and n is evaluated elsewhere in the body of f, so we can play the same
112 We must be careful not to allocate the same constructor twice. Consider
113 f p = (...(case p of (a,b) -> e)...p...,
114 ...let t = (r,s) in ...t...(f t)...)
115 At the recursive call to f, we can see that t is a pair. But we do NOT want
116 to make a specialised copy:
117 f' a b = let p = (a,b) in (..., ...)
118 because now t is allocated by the caller, then r and s are passed to the
119 recursive call, which allocates the (r,s) pair again.
122 (a) the argument p is used in other than a case-scrutinsation way.
123 (b) the argument to the call is not a 'fresh' tuple; you have to
124 look into its unfolding to see that it's a tuple
126 Hence the "OR" part of Note [Good arguments] below.
128 ALTERNATIVE 2: pass both boxed and unboxed versions. This no longer saves
129 allocation, but does perhaps save evals. In the RULE we'd have
132 f (I# x#) = f' (I# x#) x#
134 If at the call site the (I# x) was an unfolding, then we'd have to
135 rely on CSE to eliminate the duplicate allocation.... This alternative
136 doesn't look attractive enough to pursue.
138 ALTERNATIVE 3: ignore the reboxing problem. The trouble is that
139 the conservative reboxing story prevents many useful functions from being
140 specialised. Example:
141 foo :: Maybe Int -> Int -> Int
143 foo x@(Just m) n = foo x (n-m)
144 Here the use of 'x' will clearly not require boxing in the specialised function.
146 The strictness analyser has the same problem, in fact. Example:
148 If we pass just 'a' and 'b' to the worker, it might need to rebox the
149 pair to create (a,b). A more sophisticated analysis might figure out
150 precisely the cases in which this could happen, but the strictness
151 analyser does no such analysis; it just passes 'a' and 'b', and hopes
154 So my current choice is to make SpecConstr similarly aggressive, and
155 ignore the bad potential of reboxing.
158 Note [Good arguments]
159 ~~~~~~~~~~~~~~~~~~~~~
162 * A self-recursive function. Ignore mutual recursion for now,
163 because it's less common, and the code is simpler for self-recursion.
167 a) At a recursive call, one or more parameters is an explicit
168 constructor application
170 That same parameter is scrutinised by a case somewhere in
171 the RHS of the function
175 b) At a recursive call, one or more parameters has an unfolding
176 that is an explicit constructor application
178 That same parameter is scrutinised by a case somewhere in
179 the RHS of the function
181 Those are the only uses of the parameter (see Note [Reboxing])
184 What to abstract over
185 ~~~~~~~~~~~~~~~~~~~~~
186 There's a bit of a complication with type arguments. If the call
189 f p = ...f ((:) [a] x xs)...
191 then our specialised function look like
193 f_spec x xs = let p = (:) [a] x xs in ....as before....
195 This only makes sense if either
196 a) the type variable 'a' is in scope at the top of f, or
197 b) the type variable 'a' is an argument to f (and hence fs)
199 Actually, (a) may hold for value arguments too, in which case
200 we may not want to pass them. Supose 'x' is in scope at f's
201 defn, but xs is not. Then we'd like
203 f_spec xs = let p = (:) [a] x xs in ....as before....
205 Similarly (b) may hold too. If x is already an argument at the
206 call, no need to pass it again.
208 Finally, if 'a' is not in scope at the call site, we could abstract
209 it as we do the term variables:
211 f_spec a x xs = let p = (:) [a] x xs in ...as before...
213 So the grand plan is:
215 * abstract the call site to a constructor-only pattern
216 e.g. C x (D (f p) (g q)) ==> C s1 (D s2 s3)
218 * Find the free variables of the abstracted pattern
220 * Pass these variables, less any that are in scope at
221 the fn defn. But see Note [Shadowing] below.
224 NOTICE that we only abstract over variables that are not in scope,
225 so we're in no danger of shadowing variables used in "higher up"
231 In this pass we gather up usage information that may mention variables
232 that are bound between the usage site and the definition site; or (more
233 seriously) may be bound to something different at the definition site.
236 f x = letrec g y v = let x = ...
239 Since 'x' is in scope at the call site, we may make a rewrite rule that
241 RULE forall a,b. g (a,b) x = ...
242 But this rule will never match, because it's really a different 'x' at
243 the call site -- and that difference will be manifest by the time the
244 simplifier gets to it. [A worry: the simplifier doesn't *guarantee*
245 no-shadowing, so perhaps it may not be distinct?]
247 Anyway, the rule isn't actually wrong, it's just not useful. One possibility
248 is to run deShadowBinds before running SpecConstr, but instead we run the
249 simplifier. That gives the simplest possible program for SpecConstr to
250 chew on; and it virtually guarantees no shadowing.
252 Note [Specialising for constant parameters]
253 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
254 This one is about specialising on a *constant* (but not necessarily
255 constructor) argument
257 foo :: Int -> (Int -> Int) -> Int
259 foo m f = foo (f m) (+1)
263 lvl_rmV :: GHC.Base.Int -> GHC.Base.Int
265 \ (ds_dlk :: GHC.Base.Int) ->
266 case ds_dlk of wild_alH { GHC.Base.I# x_alG ->
267 GHC.Base.I# (GHC.Prim.+# x_alG 1)
269 T.$wfoo :: GHC.Prim.Int# -> (GHC.Base.Int -> GHC.Base.Int) ->
272 \ (ww_sme :: GHC.Prim.Int#) (w_smg :: GHC.Base.Int -> GHC.Base.Int) ->
273 case ww_sme of ds_Xlw {
275 case w_smg (GHC.Base.I# ds_Xlw) of w1_Xmo { GHC.Base.I# ww1_Xmz ->
276 T.$wfoo ww1_Xmz lvl_rmV
281 The recursive call has lvl_rmV as its argument, so we could create a specialised copy
282 with that argument baked in; that is, not passed at all. Now it can perhaps be inlined.
284 When is this worth it? Call the constant 'lvl'
285 - If 'lvl' has an unfolding that is a constructor, see if the corresponding
286 parameter is scrutinised anywhere in the body.
288 - If 'lvl' has an unfolding that is a inlinable function, see if the corresponding
289 parameter is applied (...to enough arguments...?)
291 Also do this is if the function has RULES?
295 Note [Specialising for lambda parameters]
296 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
297 foo :: Int -> (Int -> Int) -> Int
299 foo m f = foo (f m) (\n -> n-m)
301 This is subtly different from the previous one in that we get an
302 explicit lambda as the argument:
304 T.$wfoo :: GHC.Prim.Int# -> (GHC.Base.Int -> GHC.Base.Int) ->
307 \ (ww_sm8 :: GHC.Prim.Int#) (w_sma :: GHC.Base.Int -> GHC.Base.Int) ->
308 case ww_sm8 of ds_Xlr {
310 case w_sma (GHC.Base.I# ds_Xlr) of w1_Xmf { GHC.Base.I# ww1_Xmq ->
313 (\ (n_ad3 :: GHC.Base.Int) ->
314 case n_ad3 of wild_alB { GHC.Base.I# x_alA ->
315 GHC.Base.I# (GHC.Prim.-# x_alA ds_Xlr)
321 I wonder if SpecConstr couldn't be extended to handle this? After all,
322 lambda is a sort of constructor for functions and perhaps it already
323 has most of the necessary machinery?
325 Furthermore, there's an immediate win, because you don't need to allocate the lamda
326 at the call site; and if perchance it's called in the recursive call, then you
327 may avoid allocating it altogether. Just like for constructors.
329 Looks cool, but probably rare...but it might be easy to implement.
332 Note [SpecConstr for casts]
333 ~~~~~~~~~~~~~~~~~~~~~~~~~~~
336 data instance T Int = T Int
341 go (T n) = go (T (n-1))
343 The recursive call ends up looking like
344 go (T (I# ...) `cast` g)
345 So we want to spot the construtor application inside the cast.
346 That's why we have the Cast case in argToPat
348 Note [Local recursive groups]
349 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
350 For a *local* recursive group, we can see all the calls to the
351 function, so we seed the specialisation loop from the calls in the
352 body, not from the calls in the RHS. Consider:
354 bar m n = foo n (n,n) (n,n) (n,n) (n,n)
358 | n > 3000 = case p of { (p1,p2) -> foo (n-1) (p2,p1) q r s }
359 | n > 2000 = case q of { (q1,q2) -> foo (n-1) p (q2,q1) r s }
360 | n > 1000 = case r of { (r1,r2) -> foo (n-1) p q (r2,r1) s }
361 | otherwise = case s of { (s1,s2) -> foo (n-1) p q r (s2,s1) }
363 If we start with the RHSs of 'foo', we get lots and lots of specialisations,
364 most of which are not needed. But if we start with the (single) call
365 in the rhs of 'bar' we get exactly one fully-specialised copy, and all
366 the recursive calls go to this fully-specialised copy. Indeed, the original
367 function is later collected as dead code. This is very important in
368 specialising the loops arising from stream fusion, for example in NDP where
369 we were getting literally hundreds of (mostly unused) specialisations of
372 -----------------------------------------------------
373 Stuff not yet handled
374 -----------------------------------------------------
376 Here are notes arising from Roman's work that I don't want to lose.
382 foo :: Int -> T Int -> Int
384 foo x t | even x = case t of { T n -> foo (x-n) t }
385 | otherwise = foo (x-1) t
387 SpecConstr does no specialisation, because the second recursive call
388 looks like a boxed use of the argument. A pity.
390 $wfoo_sFw :: GHC.Prim.Int# -> T.T GHC.Base.Int -> GHC.Prim.Int#
392 \ (ww_sFo [Just L] :: GHC.Prim.Int#) (w_sFq [Just L] :: T.T GHC.Base.Int) ->
393 case ww_sFo of ds_Xw6 [Just L] {
395 case GHC.Prim.remInt# ds_Xw6 2 of wild1_aEF [Dead Just A] {
396 __DEFAULT -> $wfoo_sFw (GHC.Prim.-# ds_Xw6 1) w_sFq;
398 case w_sFq of wild_Xy [Just L] { T.T n_ad5 [Just U(L)] ->
399 case n_ad5 of wild1_aET [Just A] { GHC.Base.I# y_aES [Just L] ->
400 $wfoo_sFw (GHC.Prim.-# ds_Xw6 y_aES) wild_Xy
406 data a :*: b = !a :*: !b
409 foo :: (Int :*: T Int) -> Int
411 foo (x :*: t) | even x = case t of { T n -> foo ((x-n) :*: t) }
412 | otherwise = foo ((x-1) :*: t)
414 Very similar to the previous one, except that the parameters are now in
415 a strict tuple. Before SpecConstr, we have
417 $wfoo_sG3 :: GHC.Prim.Int# -> T.T GHC.Base.Int -> GHC.Prim.Int#
419 \ (ww_sFU [Just L] :: GHC.Prim.Int#) (ww_sFW [Just L] :: T.T
421 case ww_sFU of ds_Xws [Just L] {
423 case GHC.Prim.remInt# ds_Xws 2 of wild1_aEZ [Dead Just A] {
425 case ww_sFW of tpl_B2 [Just L] { T.T a_sFo [Just A] ->
426 $wfoo_sG3 (GHC.Prim.-# ds_Xws 1) tpl_B2 -- $wfoo1
429 case ww_sFW of wild_XB [Just A] { T.T n_ad7 [Just S(L)] ->
430 case n_ad7 of wild1_aFd [Just L] { GHC.Base.I# y_aFc [Just L] ->
431 $wfoo_sG3 (GHC.Prim.-# ds_Xws y_aFc) wild_XB -- $wfoo2
435 We get two specialisations:
436 "SC:$wfoo1" [0] __forall {a_sFB :: GHC.Base.Int sc_sGC :: GHC.Prim.Int#}
437 Foo.$wfoo sc_sGC (Foo.T @ GHC.Base.Int a_sFB)
438 = Foo.$s$wfoo1 a_sFB sc_sGC ;
439 "SC:$wfoo2" [0] __forall {y_aFp :: GHC.Prim.Int# sc_sGC :: GHC.Prim.Int#}
440 Foo.$wfoo sc_sGC (Foo.T @ GHC.Base.Int (GHC.Base.I# y_aFp))
441 = Foo.$s$wfoo y_aFp sc_sGC ;
443 But perhaps the first one isn't good. After all, we know that tpl_B2 is
444 a T (I# x) really, because T is strict and Int has one constructor. (We can't
445 unbox the strict fields, becuase T is polymorphic!)
449 %************************************************************************
451 \subsection{Top level wrapper stuff}
453 %************************************************************************
456 specConstrProgram :: DynFlags -> UniqSupply -> [CoreBind] -> IO [CoreBind]
457 specConstrProgram dflags us binds
459 showPass dflags "SpecConstr"
461 let (binds', _) = initUs us (go (initScEnv dflags) binds)
463 endPass dflags "SpecConstr" Opt_D_dump_spec binds'
465 dumpIfSet_dyn dflags Opt_D_dump_rules "Top-level specialisations"
466 (pprRules (tidyRules emptyTidyEnv (rulesOfBinds binds')))
471 go env (bind:binds) = do (env', bind') <- scTopBind env bind
472 binds' <- go env' binds
473 return (bind' : binds')
477 %************************************************************************
479 \subsection{Environment: goes downwards}
481 %************************************************************************
484 data ScEnv = SCE { sc_size :: Maybe Int, -- Size threshold
485 sc_count :: Maybe Int, -- Max # of specialisations for any one fn
487 sc_subst :: Subst, -- Current substitution
488 -- Maps InIds to OutExprs
490 sc_how_bound :: HowBoundEnv,
491 -- Binds interesting non-top-level variables
492 -- Domain is OutVars (*after* applying the substitution)
495 -- Domain is OutIds (*after* applying the substitution)
496 -- Used even for top-level bindings (but not imported ones)
499 ---------------------
500 -- As we go, we apply a substitution (sc_subst) to the current term
501 type InExpr = CoreExpr -- *Before* applying the subst
503 type OutExpr = CoreExpr -- *After* applying the subst
507 ---------------------
508 type HowBoundEnv = VarEnv HowBound -- Domain is OutVars
510 ---------------------
511 type ValueEnv = IdEnv Value -- Domain is OutIds
512 data Value = ConVal AltCon [CoreArg] -- *Saturated* constructors
513 | LambdaVal -- Inlinable lambdas or PAPs
515 instance Outputable Value where
516 ppr (ConVal con args) = ppr con <+> interpp'SP args
517 ppr LambdaVal = ptext (sLit "<Lambda>")
519 ---------------------
520 initScEnv :: DynFlags -> ScEnv
522 = SCE { sc_size = specConstrThreshold dflags,
523 sc_count = specConstrCount dflags,
524 sc_subst = emptySubst,
525 sc_how_bound = emptyVarEnv,
526 sc_vals = emptyVarEnv }
528 data HowBound = RecFun -- These are the recursive functions for which
529 -- we seek interesting call patterns
531 | RecArg -- These are those functions' arguments, or their sub-components;
532 -- we gather occurrence information for these
534 instance Outputable HowBound where
535 ppr RecFun = text "RecFun"
536 ppr RecArg = text "RecArg"
538 lookupHowBound :: ScEnv -> Id -> Maybe HowBound
539 lookupHowBound env id = lookupVarEnv (sc_how_bound env) id
541 scSubstId :: ScEnv -> Id -> CoreExpr
542 scSubstId env v = lookupIdSubst (sc_subst env) v
544 scSubstTy :: ScEnv -> Type -> Type
545 scSubstTy env ty = substTy (sc_subst env) ty
547 zapScSubst :: ScEnv -> ScEnv
548 zapScSubst env = env { sc_subst = zapSubstEnv (sc_subst env) }
550 extendScInScope :: ScEnv -> [Var] -> ScEnv
551 -- Bring the quantified variables into scope
552 extendScInScope env qvars = env { sc_subst = extendInScopeList (sc_subst env) qvars }
554 -- Extend the substitution
555 extendScSubst :: ScEnv -> Var -> OutExpr -> ScEnv
556 extendScSubst env var expr = env { sc_subst = extendSubst (sc_subst env) var expr }
558 extendScSubstList :: ScEnv -> [(Var,OutExpr)] -> ScEnv
559 extendScSubstList env prs = env { sc_subst = extendSubstList (sc_subst env) prs }
561 extendHowBound :: ScEnv -> [Var] -> HowBound -> ScEnv
562 extendHowBound env bndrs how_bound
563 = env { sc_how_bound = extendVarEnvList (sc_how_bound env)
564 [(bndr,how_bound) | bndr <- bndrs] }
566 extendBndrsWith :: HowBound -> ScEnv -> [Var] -> (ScEnv, [Var])
567 extendBndrsWith how_bound env bndrs
568 = (env { sc_subst = subst', sc_how_bound = hb_env' }, bndrs')
570 (subst', bndrs') = substBndrs (sc_subst env) bndrs
571 hb_env' = sc_how_bound env `extendVarEnvList`
572 [(bndr,how_bound) | bndr <- bndrs']
574 extendBndrWith :: HowBound -> ScEnv -> Var -> (ScEnv, Var)
575 extendBndrWith how_bound env bndr
576 = (env { sc_subst = subst', sc_how_bound = hb_env' }, bndr')
578 (subst', bndr') = substBndr (sc_subst env) bndr
579 hb_env' = extendVarEnv (sc_how_bound env) bndr' how_bound
581 extendRecBndrs :: ScEnv -> [Var] -> (ScEnv, [Var])
582 extendRecBndrs env bndrs = (env { sc_subst = subst' }, bndrs')
584 (subst', bndrs') = substRecBndrs (sc_subst env) bndrs
586 extendBndr :: ScEnv -> Var -> (ScEnv, Var)
587 extendBndr env bndr = (env { sc_subst = subst' }, bndr')
589 (subst', bndr') = substBndr (sc_subst env) bndr
591 extendValEnv :: ScEnv -> Id -> Maybe Value -> ScEnv
592 extendValEnv env _ Nothing = env
593 extendValEnv env id (Just cv) = env { sc_vals = extendVarEnv (sc_vals env) id cv }
595 extendCaseBndrs :: ScEnv -> CoreExpr -> Id -> AltCon -> [Var] -> ScEnv
599 -- we want to bind b, and perhaps scrut too, to (C x y)
600 -- NB: Extends only the sc_vals part of the envt
601 extendCaseBndrs env scrut case_bndr con alt_bndrs
603 Var v -> extendValEnv env1 v cval
606 env1 = extendValEnv env case_bndr cval
609 LitAlt {} -> Just (ConVal con [])
610 DataAlt {} -> Just (ConVal con vanilla_args)
612 vanilla_args = map Type (tyConAppArgs (idType case_bndr)) ++
613 varsToCoreExprs alt_bndrs
617 %************************************************************************
619 \subsection{Usage information: flows upwards}
621 %************************************************************************
626 scu_calls :: CallEnv, -- Calls
627 -- The functions are a subset of the
628 -- RecFuns in the ScEnv
630 scu_occs :: !(IdEnv ArgOcc) -- Information on argument occurrences
631 } -- The domain is OutIds
633 type CallEnv = IdEnv [Call]
634 type Call = (ValueEnv, [CoreArg])
635 -- The arguments of the call, together with the
636 -- env giving the constructor bindings at the call site
639 nullUsage = SCU { scu_calls = emptyVarEnv, scu_occs = emptyVarEnv }
641 combineCalls :: CallEnv -> CallEnv -> CallEnv
642 combineCalls = plusVarEnv_C (++)
644 combineUsage :: ScUsage -> ScUsage -> ScUsage
645 combineUsage u1 u2 = SCU { scu_calls = combineCalls (scu_calls u1) (scu_calls u2),
646 scu_occs = plusVarEnv_C combineOcc (scu_occs u1) (scu_occs u2) }
648 combineUsages :: [ScUsage] -> ScUsage
649 combineUsages [] = nullUsage
650 combineUsages us = foldr1 combineUsage us
652 lookupOcc :: ScUsage -> OutVar -> (ScUsage, ArgOcc)
653 lookupOcc (SCU { scu_calls = sc_calls, scu_occs = sc_occs }) bndr
654 = (SCU {scu_calls = sc_calls, scu_occs = delVarEnv sc_occs bndr},
655 lookupVarEnv sc_occs bndr `orElse` NoOcc)
657 lookupOccs :: ScUsage -> [OutVar] -> (ScUsage, [ArgOcc])
658 lookupOccs (SCU { scu_calls = sc_calls, scu_occs = sc_occs }) bndrs
659 = (SCU {scu_calls = sc_calls, scu_occs = delVarEnvList sc_occs bndrs},
660 [lookupVarEnv sc_occs b `orElse` NoOcc | b <- bndrs])
662 data ArgOcc = NoOcc -- Doesn't occur at all; or a type argument
663 | UnkOcc -- Used in some unknown way
665 | ScrutOcc (UniqFM [ArgOcc]) -- See Note [ScrutOcc]
667 | BothOcc -- Definitely taken apart, *and* perhaps used in some other way
671 An occurrence of ScrutOcc indicates that the thing, or a `cast` version of the thing,
672 is *only* taken apart or applied.
674 Functions, literal: ScrutOcc emptyUFM
675 Data constructors: ScrutOcc subs,
677 where (subs :: UniqFM [ArgOcc]) gives usage of the *pattern-bound* components,
678 The domain of the UniqFM is the Unique of the data constructor
680 The [ArgOcc] is the occurrences of the *pattern-bound* components
681 of the data structure. E.g.
682 data T a = forall b. MkT a b (b->a)
683 A pattern binds b, x::a, y::b, z::b->a, but not 'a'!
687 instance Outputable ArgOcc where
688 ppr (ScrutOcc xs) = ptext (sLit "scrut-occ") <> ppr xs
689 ppr UnkOcc = ptext (sLit "unk-occ")
690 ppr BothOcc = ptext (sLit "both-occ")
691 ppr NoOcc = ptext (sLit "no-occ")
693 -- Experimentally, this vesion of combineOcc makes ScrutOcc "win", so
694 -- that if the thing is scrutinised anywhere then we get to see that
695 -- in the overall result, even if it's also used in a boxed way
696 -- This might be too agressive; see Note [Reboxing] Alternative 3
697 combineOcc :: ArgOcc -> ArgOcc -> ArgOcc
698 combineOcc NoOcc occ = occ
699 combineOcc occ NoOcc = occ
700 combineOcc (ScrutOcc xs) (ScrutOcc ys) = ScrutOcc (plusUFM_C combineOccs xs ys)
701 combineOcc _occ (ScrutOcc ys) = ScrutOcc ys
702 combineOcc (ScrutOcc xs) _occ = ScrutOcc xs
703 combineOcc UnkOcc UnkOcc = UnkOcc
704 combineOcc _ _ = BothOcc
706 combineOccs :: [ArgOcc] -> [ArgOcc] -> [ArgOcc]
707 combineOccs xs ys = zipWithEqual "combineOccs" combineOcc xs ys
709 setScrutOcc :: ScEnv -> ScUsage -> OutExpr -> ArgOcc -> ScUsage
710 -- *Overwrite* the occurrence info for the scrutinee, if the scrutinee
711 -- is a variable, and an interesting variable
712 setScrutOcc env usg (Cast e _) occ = setScrutOcc env usg e occ
713 setScrutOcc env usg (Note _ e) occ = setScrutOcc env usg e occ
714 setScrutOcc env usg (Var v) occ
715 | Just RecArg <- lookupHowBound env v = usg { scu_occs = extendVarEnv (scu_occs usg) v occ }
717 setScrutOcc _env usg _other _occ -- Catch-all
720 conArgOccs :: ArgOcc -> AltCon -> [ArgOcc]
721 -- Find usage of components of data con; returns [UnkOcc...] if unknown
722 -- See Note [ScrutOcc] for the extra UnkOccs in the vanilla datacon case
724 conArgOccs (ScrutOcc fm) (DataAlt dc)
725 | Just pat_arg_occs <- lookupUFM fm dc
726 = [UnkOcc | _ <- dataConUnivTyVars dc] ++ pat_arg_occs
728 conArgOccs _other _con = repeat UnkOcc
731 %************************************************************************
733 \subsection{The main recursive function}
735 %************************************************************************
737 The main recursive function gathers up usage information, and
738 creates specialised versions of functions.
741 scExpr, scExpr' :: ScEnv -> CoreExpr -> UniqSM (ScUsage, CoreExpr)
742 -- The unique supply is needed when we invent
743 -- a new name for the specialised function and its args
745 scExpr env e = scExpr' env e
748 scExpr' env (Var v) = case scSubstId env v of
749 Var v' -> return (varUsage env v' UnkOcc, Var v')
750 e' -> scExpr (zapScSubst env) e'
752 scExpr' env (Type t) = return (nullUsage, Type (scSubstTy env t))
753 scExpr' _ e@(Lit {}) = return (nullUsage, e)
754 scExpr' env (Note n e) = do (usg,e') <- scExpr env e
755 return (usg, Note n e')
756 scExpr' env (Cast e co) = do (usg, e') <- scExpr env e
757 return (usg, Cast e' (scSubstTy env co))
758 scExpr' env e@(App _ _) = scApp env (collectArgs e)
759 scExpr' env (Lam b e) = do let (env', b') = extendBndr env b
760 (usg, e') <- scExpr env' e
761 return (usg, Lam b' e')
763 scExpr' env (Case scrut b ty alts)
764 = do { (scrut_usg, scrut') <- scExpr env scrut
765 ; case isValue (sc_vals env) scrut' of
766 Just (ConVal con args) -> sc_con_app con args scrut'
767 _other -> sc_vanilla scrut_usg scrut'
770 sc_con_app con args scrut' -- Known constructor; simplify
771 = do { let (_, bs, rhs) = findAlt con alts
772 alt_env' = extendScSubstList env ((b,scrut') : bs `zip` trimConArgs con args)
773 ; scExpr alt_env' rhs }
775 sc_vanilla scrut_usg scrut' -- Normal case
776 = do { let (alt_env,b') = extendBndrWith RecArg env b
777 -- Record RecArg for the components
779 ; (alt_usgs, alt_occs, alts')
780 <- mapAndUnzip3M (sc_alt alt_env scrut' b') alts
782 ; let (alt_usg, b_occ) = lookupOcc (combineUsages alt_usgs) b'
783 scrut_occ = foldr combineOcc b_occ alt_occs
784 scrut_usg' = setScrutOcc env scrut_usg scrut' scrut_occ
785 -- The combined usage of the scrutinee is given
786 -- by scrut_occ, which is passed to scScrut, which
787 -- in turn treats a bare-variable scrutinee specially
789 ; return (alt_usg `combineUsage` scrut_usg',
790 Case scrut' b' (scSubstTy env ty) alts') }
792 sc_alt env scrut' b' (con,bs,rhs)
793 = do { let (env1, bs') = extendBndrsWith RecArg env bs
794 env2 = extendCaseBndrs env1 scrut' b' con bs'
795 ; (usg,rhs') <- scExpr env2 rhs
796 ; let (usg', arg_occs) = lookupOccs usg bs'
797 scrut_occ = case con of
798 DataAlt dc -> ScrutOcc (unitUFM dc arg_occs)
799 _ -> ScrutOcc emptyUFM
800 ; return (usg', scrut_occ, (con,bs',rhs')) }
802 scExpr' env (Let (NonRec bndr rhs) body)
803 | isTyVar bndr -- Type-lets may be created by doBeta
804 = scExpr' (extendScSubst env bndr rhs) body
806 = do { let (body_env, bndr') = extendBndr env bndr
807 ; (rhs_usg, (_, args', rhs_body', _)) <- scRecRhs env (bndr',rhs)
808 ; let rhs' = mkLams args' rhs_body'
810 ; if not opt_SpecInlineJoinPoints || null args' || isEmptyVarEnv (scu_calls rhs_usg) then do
812 let body_env2 = extendValEnv body_env bndr' (isValue (sc_vals env) rhs')
813 -- Record if the RHS is a value
814 ; (body_usg, body') <- scExpr body_env2 body
815 ; return (body_usg `combineUsage` rhs_usg, Let (NonRec bndr' rhs') body') }
816 else -- For now, just brutally inline the join point
817 do { let body_env2 = extendScSubst env bndr rhs'
818 ; scExpr body_env2 body } }
822 do { -- Join-point case
823 let body_env2 = extendHowBound body_env [bndr'] RecFun
824 -- If the RHS of this 'let' contains calls
825 -- to recursive functions that we're trying
826 -- to specialise, then treat this let too
827 -- as one to specialise
828 ; (body_usg, body') <- scExpr body_env2 body
830 ; (spec_usg, _, specs) <- specialise env (scu_calls body_usg) ([], rhs_info)
832 ; return (body_usg { scu_calls = scu_calls body_usg `delVarEnv` bndr' }
833 `combineUsage` rhs_usg `combineUsage` spec_usg,
834 mkLets [NonRec b r | (b,r) <- specInfoBinds rhs_info specs] body')
838 -- A *local* recursive group: see Note [Local recursive groups]
839 scExpr' env (Let (Rec prs) body)
840 = do { let (bndrs,rhss) = unzip prs
841 (rhs_env1,bndrs') = extendRecBndrs env bndrs
842 rhs_env2 = extendHowBound rhs_env1 bndrs' RecFun
844 ; (rhs_usgs, rhs_infos) <- mapAndUnzipM (scRecRhs rhs_env2) (bndrs' `zip` rhss)
845 ; (body_usg, body') <- scExpr rhs_env2 body
847 -- NB: start specLoop from body_usg
848 ; (spec_usg, specs) <- specLoop rhs_env2 (scu_calls body_usg) rhs_infos nullUsage
849 [SI [] 0 (Just usg) | usg <- rhs_usgs]
851 ; let all_usg = spec_usg `combineUsage` body_usg
852 bind' = Rec (concat (zipWith specInfoBinds rhs_infos specs))
854 ; return (all_usg { scu_calls = scu_calls all_usg `delVarEnvList` bndrs' },
857 -----------------------------------
858 scApp :: ScEnv -> (InExpr, [InExpr]) -> UniqSM (ScUsage, CoreExpr)
860 scApp env (Var fn, args) -- Function is a variable
861 = ASSERT( not (null args) )
862 do { args_w_usgs <- mapM (scExpr env) args
863 ; let (arg_usgs, args') = unzip args_w_usgs
864 arg_usg = combineUsages arg_usgs
865 ; case scSubstId env fn of
866 fn'@(Lam {}) -> scExpr (zapScSubst env) (doBeta fn' args')
867 -- Do beta-reduction and try again
869 Var fn' -> return (arg_usg `combineUsage` fn_usg, mkApps (Var fn') args')
871 fn_usg = case lookupHowBound env fn' of
872 Just RecFun -> SCU { scu_calls = unitVarEnv fn' [(sc_vals env, args')],
873 scu_occs = emptyVarEnv }
874 Just RecArg -> SCU { scu_calls = emptyVarEnv,
875 scu_occs = unitVarEnv fn' (ScrutOcc emptyUFM) }
879 other_fn' -> return (arg_usg, mkApps other_fn' args') }
880 -- NB: doing this ignores any usage info from the substituted
881 -- function, but I don't think that matters. If it does
884 doBeta :: OutExpr -> [OutExpr] -> OutExpr
885 -- ToDo: adjust for System IF
886 doBeta (Lam bndr body) (arg : args) = Let (NonRec bndr arg) (doBeta body args)
887 doBeta fn args = mkApps fn args
889 -- The function is almost always a variable, but not always.
890 -- In particular, if this pass follows float-in,
891 -- which it may, we can get
892 -- (let f = ...f... in f) arg1 arg2
893 scApp env (other_fn, args)
894 = do { (fn_usg, fn') <- scExpr env other_fn
895 ; (arg_usgs, args') <- mapAndUnzipM (scExpr env) args
896 ; return (combineUsages arg_usgs `combineUsage` fn_usg, mkApps fn' args') }
898 ----------------------
899 scTopBind :: ScEnv -> CoreBind -> UniqSM (ScEnv, CoreBind)
900 scTopBind env (Rec prs)
901 | Just threshold <- sc_size env
902 , not (all (couldBeSmallEnoughToInline threshold) rhss)
904 = do { let (rhs_env,bndrs') = extendRecBndrs env bndrs
905 ; (_, rhss') <- mapAndUnzipM (scExpr rhs_env) rhss
906 ; return (rhs_env, Rec (bndrs' `zip` rhss')) }
907 | otherwise -- Do specialisation
908 = do { let (rhs_env1,bndrs') = extendRecBndrs env bndrs
909 rhs_env2 = extendHowBound rhs_env1 bndrs' RecFun
911 ; (rhs_usgs, rhs_infos) <- mapAndUnzipM (scRecRhs rhs_env2) (bndrs' `zip` rhss)
912 ; let rhs_usg = combineUsages rhs_usgs
914 ; (_, specs) <- specLoop rhs_env2 (scu_calls rhs_usg) rhs_infos nullUsage
915 [SI [] 0 Nothing | _ <- bndrs]
917 ; return (rhs_env1, -- For the body of the letrec, delete the RecFun business
918 Rec (concat (zipWith specInfoBinds rhs_infos specs))) }
920 (bndrs,rhss) = unzip prs
922 scTopBind env (NonRec bndr rhs)
923 = do { (_, rhs') <- scExpr env rhs
924 ; let (env1, bndr') = extendBndr env bndr
925 env2 = extendValEnv env1 bndr' (isValue (sc_vals env) rhs')
926 ; return (env2, NonRec bndr' rhs') }
928 ----------------------
929 scRecRhs :: ScEnv -> (OutId, InExpr) -> UniqSM (ScUsage, RhsInfo)
930 scRecRhs env (bndr,rhs)
931 = do { let (arg_bndrs,body) = collectBinders rhs
932 (body_env, arg_bndrs') = extendBndrsWith RecArg env arg_bndrs
933 ; (body_usg, body') <- scExpr body_env body
934 ; let (rhs_usg, arg_occs) = lookupOccs body_usg arg_bndrs'
935 ; return (rhs_usg, (bndr, arg_bndrs', body', arg_occs)) }
937 -- The arg_occs says how the visible,
938 -- lambda-bound binders of the RHS are used
939 -- (including the TyVar binders)
940 -- Two pats are the same if they match both ways
942 ----------------------
943 specInfoBinds :: RhsInfo -> SpecInfo -> [(Id,CoreExpr)]
944 specInfoBinds (fn, args, body, _) (SI specs _ _)
945 = [(id,rhs) | OS _ _ id rhs <- specs] ++
946 [(fn `addIdSpecialisations` rules, mkLams args body)]
948 rules = [r | OS _ r _ _ <- specs]
950 ----------------------
951 varUsage :: ScEnv -> OutVar -> ArgOcc -> ScUsage
953 | Just RecArg <- lookupHowBound env v = SCU { scu_calls = emptyVarEnv
954 , scu_occs = unitVarEnv v use }
955 | otherwise = nullUsage
959 %************************************************************************
961 The specialiser itself
963 %************************************************************************
966 type RhsInfo = (OutId, [OutVar], OutExpr, [ArgOcc])
967 -- Info about the *original* RHS of a binding we are specialising
968 -- Original binding f = \xs.body
969 -- Plus info about usage of arguments
971 data SpecInfo = SI [OneSpec] -- The specialisations we have generated
972 Int -- Length of specs; used for numbering them
973 (Maybe ScUsage) -- Nothing => we have generated specialisations
974 -- from calls in the *original* RHS
975 -- Just cs => we haven't, and this is the usage
976 -- of the original RHS
978 -- One specialisation: Rule plus definition
979 data OneSpec = OS CallPat -- Call pattern that generated this specialisation
980 CoreRule -- Rule connecting original id with the specialisation
981 OutId OutExpr -- Spec id + its rhs
987 -> ScUsage -> [SpecInfo] -- One per binder; acccumulating parameter
988 -> UniqSM (ScUsage, [SpecInfo]) -- ...ditto...
989 specLoop env all_calls rhs_infos usg_so_far specs_so_far
990 = do { specs_w_usg <- zipWithM (specialise env all_calls) rhs_infos specs_so_far
991 ; let (new_usg_s, all_specs) = unzip specs_w_usg
992 new_usg = combineUsages new_usg_s
993 new_calls = scu_calls new_usg
994 all_usg = usg_so_far `combineUsage` new_usg
995 ; if isEmptyVarEnv new_calls then
996 return (all_usg, all_specs)
998 specLoop env new_calls rhs_infos all_usg all_specs }
1002 -> CallEnv -- Info on calls
1004 -> SpecInfo -- Original RHS plus patterns dealt with
1005 -> UniqSM (ScUsage, SpecInfo) -- New specialised versions and their usage
1007 -- Note: the rhs here is the optimised version of the original rhs
1008 -- So when we make a specialised copy of the RHS, we're starting
1009 -- from an RHS whose nested functions have been optimised already.
1011 specialise env bind_calls (fn, arg_bndrs, body, arg_occs)
1012 spec_info@(SI specs spec_count mb_unspec)
1013 | notNull arg_bndrs, -- Only specialise functions
1014 Just all_calls <- lookupVarEnv bind_calls fn
1015 = do { (boring_call, pats) <- callsToPats env specs arg_occs all_calls
1016 -- ; pprTrace "specialise" (vcat [ppr fn <+> ppr arg_occs,
1017 -- text "calls" <+> ppr all_calls,
1018 -- text "good pats" <+> ppr pats]) $
1021 -- Bale out if too many specialisations
1022 -- Rather a hacky way to do so, but it'll do for now
1023 ; let spec_count' = length pats + spec_count
1024 ; case sc_count env of
1025 Just max | spec_count' > max
1026 -> pprTrace "SpecConstr: too many specialisations for one function (see -fspec-constr-count):"
1027 (vcat [ptext (sLit "Function:") <+> ppr fn,
1028 ptext (sLit "Specialisations:") <+> ppr (pats ++ [p | OS p _ _ _ <- specs])])
1029 return (nullUsage, spec_info)
1033 { (spec_usgs, new_specs) <- mapAndUnzipM (spec_one env fn arg_bndrs body)
1034 (pats `zip` [spec_count..])
1036 ; let spec_usg = combineUsages spec_usgs
1037 (new_usg, mb_unspec')
1039 Just rhs_usg | boring_call -> (spec_usg `combineUsage` rhs_usg, Nothing)
1040 _ -> (spec_usg, mb_unspec)
1042 ; return (new_usg, SI (new_specs ++ specs) spec_count' mb_unspec') } }
1044 = return (nullUsage, spec_info) -- The boring case
1047 ---------------------
1049 -> OutId -- Function
1050 -> [Var] -- Lambda-binders of RHS; should match patterns
1051 -> CoreExpr -- Body of the original function
1053 -> UniqSM (ScUsage, OneSpec) -- Rule and binding
1055 -- spec_one creates a specialised copy of the function, together
1056 -- with a rule for using it. I'm very proud of how short this
1057 -- function is, considering what it does :-).
1063 f = /\b \y::[(a,b)] -> ....f (b,c) ((:) (a,(b,c)) (x,v) (h w))...
1064 [c::*, v::(b,c) are presumably bound by the (...) part]
1066 f_spec = /\ b c \ v::(b,c) hw::[(a,(b,c))] ->
1067 (...entire body of f...) [b -> (b,c),
1068 y -> ((:) (a,(b,c)) (x,v) hw)]
1070 RULE: forall b::* c::*, -- Note, *not* forall a, x
1074 f (b,c) ((:) (a,(b,c)) (x,v) hw) = f_spec b c v hw
1077 spec_one env fn arg_bndrs body (call_pat@(qvars, pats), rule_number)
1078 = do { -- Specialise the body
1079 let spec_env = extendScSubstList (extendScInScope env qvars)
1080 (arg_bndrs `zip` pats)
1081 ; (spec_usg, spec_body) <- scExpr spec_env body
1083 -- ; pprTrace "spec_one" (ppr fn <+> vcat [text "pats" <+> ppr pats,
1084 -- text "calls" <+> (ppr (scu_calls spec_usg))])
1087 -- And build the results
1088 ; spec_uniq <- getUniqueUs
1089 ; let (spec_lam_args, spec_call_args) = mkWorkerArgs qvars body_ty
1090 -- Usual w/w hack to avoid generating
1091 -- a spec_rhs of unlifted type and no args
1094 fn_loc = nameSrcSpan fn_name
1095 spec_occ = mkSpecOcc (nameOccName fn_name)
1096 rule_name = mkFastString ("SC:" ++ showSDoc (ppr fn <> int rule_number))
1097 spec_rhs = mkLams spec_lam_args spec_body
1098 spec_id = mkUserLocal spec_occ spec_uniq (mkPiTypes spec_lam_args body_ty) fn_loc
1099 body_ty = exprType spec_body
1100 rule_rhs = mkVarApps (Var spec_id) spec_call_args
1101 rule = mkLocalRule rule_name specConstrActivation fn_name qvars pats rule_rhs
1102 ; return (spec_usg, OS call_pat rule spec_id spec_rhs) }
1104 -- In which phase should the specialise-constructor rules be active?
1105 -- Originally I made them always-active, but Manuel found that
1106 -- this defeated some clever user-written rules. So Plan B
1107 -- is to make them active only in Phase 0; after all, currently,
1108 -- the specConstr transformation is only run after the simplifier
1109 -- has reached Phase 0. In general one would want it to be
1110 -- flag-controllable, but for now I'm leaving it baked in
1112 specConstrActivation :: Activation
1113 specConstrActivation = ActiveAfter 0 -- Baked in; see comments above
1116 %************************************************************************
1118 \subsection{Argument analysis}
1120 %************************************************************************
1122 This code deals with analysing call-site arguments to see whether
1123 they are constructor applications.
1127 type CallPat = ([Var], [CoreExpr]) -- Quantified variables and arguments
1130 callsToPats :: ScEnv -> [OneSpec] -> [ArgOcc] -> [Call] -> UniqSM (Bool, [CallPat])
1131 -- Result has no duplicate patterns,
1132 -- nor ones mentioned in done_pats
1133 -- Bool indicates that there was at least one boring pattern
1134 callsToPats env done_specs bndr_occs calls
1135 = do { mb_pats <- mapM (callToPats env bndr_occs) calls
1137 ; let good_pats :: [([Var], [CoreArg])]
1138 good_pats = catMaybes mb_pats
1139 done_pats = [p | OS p _ _ _ <- done_specs]
1140 is_done p = any (samePat p) done_pats
1142 ; return (any isNothing mb_pats,
1143 filterOut is_done (nubBy samePat good_pats)) }
1145 callToPats :: ScEnv -> [ArgOcc] -> Call -> UniqSM (Maybe CallPat)
1146 -- The [Var] is the variables to quantify over in the rule
1147 -- Type variables come first, since they may scope
1148 -- over the following term variables
1149 -- The [CoreExpr] are the argument patterns for the rule
1150 callToPats env bndr_occs (con_env, args)
1151 | length args < length bndr_occs -- Check saturated
1154 = do { let in_scope = substInScope (sc_subst env)
1155 ; prs <- argsToPats in_scope con_env (args `zip` bndr_occs)
1156 ; let (interesting_s, pats) = unzip prs
1157 pat_fvs = varSetElems (exprsFreeVars pats)
1158 qvars = filterOut (`elemInScopeSet` in_scope) pat_fvs
1159 -- Quantify over variables that are not in sccpe
1161 -- See Note [Shadowing] at the top
1163 (tvs, ids) = partition isTyVar qvars
1165 -- Put the type variables first; the type of a term
1166 -- variable may mention a type variable
1168 ; -- pprTrace "callToPats" (ppr args $$ ppr prs $$ ppr bndr_occs) $
1170 then return (Just (qvars', pats))
1171 else return Nothing }
1173 -- argToPat takes an actual argument, and returns an abstracted
1174 -- version, consisting of just the "constructor skeleton" of the
1175 -- argument, with non-constructor sub-expression replaced by new
1176 -- placeholder variables. For example:
1177 -- C a (D (f x) (g y)) ==> C p1 (D p2 p3)
1179 argToPat :: InScopeSet -- What's in scope at the fn defn site
1180 -> ValueEnv -- ValueEnv at the call site
1181 -> CoreArg -- A call arg (or component thereof)
1183 -> UniqSM (Bool, CoreArg)
1184 -- Returns (interesting, pat),
1185 -- where pat is the pattern derived from the argument
1186 -- intersting=True if the pattern is non-trivial (not a variable or type)
1187 -- E.g. x:xs --> (True, x:xs)
1188 -- f xs --> (False, w) where w is a fresh wildcard
1189 -- (f xs, 'c') --> (True, (w, 'c')) where w is a fresh wildcard
1190 -- \x. x+y --> (True, \x. x+y)
1191 -- lvl7 --> (True, lvl7) if lvl7 is bound
1192 -- somewhere further out
1194 argToPat _in_scope _val_env arg@(Type {}) _arg_occ
1195 = return (False, arg)
1197 argToPat in_scope val_env (Note _ arg) arg_occ
1198 = argToPat in_scope val_env arg arg_occ
1199 -- Note [Notes in call patterns]
1200 -- ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1201 -- Ignore Notes. In particular, we want to ignore any InlineMe notes
1202 -- Perhaps we should not ignore profiling notes, but I'm going to
1203 -- ride roughshod over them all for now.
1204 --- See Note [Notes in RULE matching] in Rules
1206 argToPat in_scope val_env (Let _ arg) arg_occ
1207 = argToPat in_scope val_env arg arg_occ
1208 -- Look through let expressions
1209 -- e.g. f (let v = rhs in \y -> ...v...)
1210 -- Here we can specialise for f (\y -> ...)
1211 -- because the rule-matcher will look through the let.
1213 argToPat in_scope val_env (Cast arg co) arg_occ
1214 = do { (interesting, arg') <- argToPat in_scope val_env arg arg_occ
1215 ; let (ty1,ty2) = coercionKind co
1216 ; if not interesting then
1219 { -- Make a wild-card pattern for the coercion
1221 ; let co_name = mkSysTvName uniq (fsLit "sg")
1222 co_var = mkCoVar co_name (mkCoKind ty1 ty2)
1223 ; return (interesting, Cast arg' (mkTyVarTy co_var)) } }
1225 {- Disabling lambda specialisation for now
1226 It's fragile, and the spec_loop can be infinite
1227 argToPat in_scope val_env arg arg_occ
1229 = return (True, arg)
1231 is_value_lam (Lam v e) -- Spot a value lambda, even if
1232 | isId v = True -- it is inside a type lambda
1233 | otherwise = is_value_lam e
1234 is_value_lam other = False
1237 -- Check for a constructor application
1238 -- NB: this *precedes* the Var case, so that we catch nullary constrs
1239 argToPat in_scope val_env arg arg_occ
1240 | Just (ConVal dc args) <- isValue val_env arg
1242 ScrutOcc _ -> True -- Used only by case scrutinee
1243 BothOcc -> case arg of -- Used elsewhere
1244 App {} -> True -- see Note [Reboxing]
1246 _other -> False -- No point; the arg is not decomposed
1247 = do { args' <- argsToPats in_scope val_env (args `zip` conArgOccs arg_occ dc)
1248 ; return (True, mk_con_app dc (map snd args')) }
1250 -- Check if the argument is a variable that
1251 -- is in scope at the function definition site
1252 -- It's worth specialising on this if
1253 -- (a) it's used in an interesting way in the body
1254 -- (b) we know what its value is
1255 argToPat in_scope val_env (Var v) arg_occ
1256 | case arg_occ of { UnkOcc -> False; _other -> True }, -- (a)
1258 = return (True, Var v)
1261 | isLocalId v = v `elemInScopeSet` in_scope
1262 && isJust (lookupVarEnv val_env v)
1263 -- Local variables have values in val_env
1264 | otherwise = isValueUnfolding (idUnfolding v)
1265 -- Imports have unfoldings
1267 -- I'm really not sure what this comment means
1268 -- And by not wild-carding we tend to get forall'd
1269 -- variables that are in soope, which in turn can
1270 -- expose the weakness in let-matching
1271 -- See Note [Matching lets] in Rules
1273 -- Check for a variable bound inside the function.
1274 -- Don't make a wild-card, because we may usefully share
1275 -- e.g. f a = let x = ... in f (x,x)
1276 -- NB: this case follows the lambda and con-app cases!!
1277 -- argToPat _in_scope _val_env (Var v) _arg_occ
1278 -- = return (False, Var v)
1279 -- SLPJ : disabling this to avoid proliferation of versions
1280 -- also works badly when thinking about seeding the loop
1281 -- from the body of the let
1282 -- f x y = letrec g z = ... in g (x,y)
1283 -- We don't want to specialise for that *particular* x,y
1285 -- The default case: make a wild-card
1286 argToPat _in_scope _val_env arg _arg_occ
1287 = wildCardPat (exprType arg)
1289 wildCardPat :: Type -> UniqSM (Bool, CoreArg)
1290 wildCardPat ty = do { uniq <- getUniqueUs
1291 ; let id = mkSysLocal (fsLit "sc") uniq ty
1292 ; return (False, Var id) }
1294 argsToPats :: InScopeSet -> ValueEnv
1295 -> [(CoreArg, ArgOcc)]
1296 -> UniqSM [(Bool, CoreArg)]
1297 argsToPats in_scope val_env args
1300 do_one (arg,occ) = argToPat in_scope val_env arg occ
1305 isValue :: ValueEnv -> CoreExpr -> Maybe Value
1306 isValue _env (Lit lit)
1307 = Just (ConVal (LitAlt lit) [])
1310 | Just stuff <- lookupVarEnv env v
1311 = Just stuff -- You might think we could look in the idUnfolding here
1312 -- but that doesn't take account of which branch of a
1313 -- case we are in, which is the whole point
1315 | not (isLocalId v) && isCheapUnfolding unf
1316 = isValue env (unfoldingTemplate unf)
1319 -- However we do want to consult the unfolding
1320 -- as well, for let-bound constructors!
1322 isValue env (Lam b e)
1323 | isTyVar b = case isValue env e of
1324 Just _ -> Just LambdaVal
1326 | otherwise = Just LambdaVal
1328 isValue _env expr -- Maybe it's a constructor application
1329 | (Var fun, args) <- collectArgs expr
1330 = case isDataConWorkId_maybe fun of
1332 Just con | args `lengthAtLeast` dataConRepArity con
1333 -- Check saturated; might be > because the
1334 -- arity excludes type args
1335 -> Just (ConVal (DataAlt con) args)
1337 _other | valArgCount args < idArity fun
1338 -- Under-applied function
1339 -> Just LambdaVal -- Partial application
1343 isValue _env _expr = Nothing
1345 mk_con_app :: AltCon -> [CoreArg] -> CoreExpr
1346 mk_con_app (LitAlt lit) [] = Lit lit
1347 mk_con_app (DataAlt con) args = mkConApp con args
1348 mk_con_app _other _args = panic "SpecConstr.mk_con_app"
1350 samePat :: CallPat -> CallPat -> Bool
1351 samePat (vs1, as1) (vs2, as2)
1354 same (Var v1) (Var v2)
1355 | v1 `elem` vs1 = v2 `elem` vs2
1356 | v2 `elem` vs2 = False
1357 | otherwise = v1 == v2
1359 same (Lit l1) (Lit l2) = l1==l2
1360 same (App f1 a1) (App f2 a2) = same f1 f2 && same a1 a2
1362 same (Type {}) (Type {}) = True -- Note [Ignore type differences]
1363 same (Note _ e1) e2 = same e1 e2 -- Ignore casts and notes
1364 same (Cast e1 _) e2 = same e1 e2
1365 same e1 (Note _ e2) = same e1 e2
1366 same e1 (Cast e2 _) = same e1 e2
1368 same e1 e2 = WARN( bad e1 || bad e2, ppr e1 $$ ppr e2)
1369 False -- Let, lambda, case should not occur
1370 bad (Case {}) = True
1376 Note [Ignore type differences]
1377 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1378 We do not want to generate specialisations where the call patterns
1379 differ only in their type arguments! Not only is it utterly useless,
1380 but it also means that (with polymorphic recursion) we can generate
1381 an infinite number of specialisations. Example is Data.Sequence.adjustTree,