2 % (c) The GRASP/AQUA Project, Glasgow University, 1992-1998
4 \section[SpecConstr]{Specialise over constructors}
11 #include "HsVersions.h"
14 import CoreLint ( showPass, endPass )
15 import CoreUtils ( exprType, mkPiTypes )
16 import CoreFVs ( exprsFreeVars )
17 import CoreSubst ( Subst, mkSubst, substExpr )
18 import CoreTidy ( tidyRules )
19 import PprCore ( pprRules )
20 import WwLib ( mkWorkerArgs )
21 import DataCon ( dataConRepArity, dataConTyVars )
22 import Type ( Type, tyConAppArgs, tyVarsOfTypes )
23 import Rules ( matchN )
24 import Id ( Id, idName, idType, isDataConWorkId_maybe,
25 mkUserLocal, mkSysLocal, idUnfolding, isLocalId )
29 import Name ( nameOccName, nameSrcLoc )
30 import Rules ( addIdSpecialisations, mkLocalRule, rulesOfBinds )
31 import OccName ( mkSpecOcc )
32 import ErrUtils ( dumpIfSet_dyn )
33 import DynFlags ( DynFlags, DynFlag(..) )
34 import BasicTypes ( Activation(..) )
35 import Maybes ( orElse, catMaybes, isJust )
36 import Util ( zipWithEqual, lengthAtLeast, notNull )
37 import List ( nubBy, partition )
44 -----------------------------------------------------
46 -----------------------------------------------------
51 drop n (x:xs) = drop (n-1) xs
53 After the first time round, we could pass n unboxed. This happens in
54 numerical code too. Here's what it looks like in Core:
56 drop n xs = case xs of
61 _ -> drop (I# (n# -# 1#)) xs
63 Notice that the recursive call has an explicit constructor as argument.
64 Noticing this, we can make a specialised version of drop
66 RULE: drop (I# n#) xs ==> drop' n# xs
68 drop' n# xs = let n = I# n# in ...orig RHS...
70 Now the simplifier will apply the specialisation in the rhs of drop', giving
72 drop' n# xs = case xs of
76 _ -> drop (n# -# 1#) xs
80 We'd also like to catch cases where a parameter is carried along unchanged,
81 but evaluated each time round the loop:
83 f i n = if i>0 || i>n then i else f (i*2) n
85 Here f isn't strict in n, but we'd like to avoid evaluating it each iteration.
86 In Core, by the time we've w/wd (f is strict in i) we get
88 f i# n = case i# ># 0 of
90 True -> case n of n' { I# n# ->
93 True -> f (i# *# 2#) n'
95 At the call to f, we see that the argument, n is know to be (I# n#),
96 and n is evaluated elsewhere in the body of f, so we can play the same
102 We must be careful not to allocate the same constructor twice. Consider
103 f p = (...(case p of (a,b) -> e)...p...,
104 ...let t = (r,s) in ...t...(f t)...)
105 At the recursive call to f, we can see that t is a pair. But we do NOT want
106 to make a specialised copy:
107 f' a b = let p = (a,b) in (..., ...)
108 because now t is allocated by the caller, then r and s are passed to the
109 recursive call, which allocates the (r,s) pair again.
112 (a) the argument p is used in other than a case-scrutinsation way.
113 (b) the argument to the call is not a 'fresh' tuple; you have to
114 look into its unfolding to see that it's a tuple
116 Hence the "OR" part of Note [Good arguments] below.
118 ALTERNATIVE: pass both boxed and unboxed versions. This no longer saves
119 allocation, but does perhaps save evals. In the RULE we'd have
122 f (I# x#) = f' (I# x#) x#
124 If at the call site the (I# x) was an unfolding, then we'd have to
125 rely on CSE to eliminate the duplicate allocation.... This alternative
126 doesn't look attractive enough to pursue.
129 Note [Good arguments]
130 ~~~~~~~~~~~~~~~~~~~~~
133 * A self-recursive function. Ignore mutual recursion for now,
134 because it's less common, and the code is simpler for self-recursion.
138 a) At a recursive call, one or more parameters is an explicit
139 constructor application
141 That same parameter is scrutinised by a case somewhere in
142 the RHS of the function
146 b) At a recursive call, one or more parameters has an unfolding
147 that is an explicit constructor application
149 That same parameter is scrutinised by a case somewhere in
150 the RHS of the function
152 Those are the only uses of the parameter (see Note [Reboxing])
155 What to abstract over
156 ~~~~~~~~~~~~~~~~~~~~~
157 There's a bit of a complication with type arguments. If the call
160 f p = ...f ((:) [a] x xs)...
162 then our specialised function look like
164 f_spec x xs = let p = (:) [a] x xs in ....as before....
166 This only makes sense if either
167 a) the type variable 'a' is in scope at the top of f, or
168 b) the type variable 'a' is an argument to f (and hence fs)
170 Actually, (a) may hold for value arguments too, in which case
171 we may not want to pass them. Supose 'x' is in scope at f's
172 defn, but xs is not. Then we'd like
174 f_spec xs = let p = (:) [a] x xs in ....as before....
176 Similarly (b) may hold too. If x is already an argument at the
177 call, no need to pass it again.
179 Finally, if 'a' is not in scope at the call site, we could abstract
180 it as we do the term variables:
182 f_spec a x xs = let p = (:) [a] x xs in ...as before...
184 So the grand plan is:
186 * abstract the call site to a constructor-only pattern
187 e.g. C x (D (f p) (g q)) ==> C s1 (D s2 s3)
189 * Find the free variables of the abstracted pattern
191 * Pass these variables, less any that are in scope at
192 the fn defn. But see Note [Shadowing] below.
195 NOTICE that we only abstract over variables that are not in scope,
196 so we're in no danger of shadowing variables used in "higher up"
202 In this pass we gather up usage information that may mention variables
203 that are bound between the usage site and the definition site; or (more
204 seriously) may be bound to something different at the definition site.
207 f x = letrec g y v = let x = ...
210 Since 'x' is in scope at the call site, we may make a rewrite rule that
212 RULE forall a,b. g (a,b) x = ...
213 But this rule will never match, because it's really a different 'x' at
214 the call site -- and that difference will be manifest by the time the
215 simplifier gets to it. [A worry: the simplifier doesn't *guarantee*
216 no-shadowing, so perhaps it may not be distinct?]
218 Anyway, the rule isn't actually wrong, it's just not useful. One possibility
219 is to run deShadowBinds before running SpecConstr, but instead we run the
220 simplifier. That gives the simplest possible program for SpecConstr to
221 chew on; and it virtually guarantees no shadowing.
223 Note [Specialising for constant parameters]
224 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
225 This one is about specialising on a *constant* (but not necessarily
226 constructor) argument
228 foo :: Int -> (Int -> Int) -> Int
230 foo m f = foo (f m) (+1)
234 lvl_rmV :: GHC.Base.Int -> GHC.Base.Int
236 \ (ds_dlk :: GHC.Base.Int) ->
237 case ds_dlk of wild_alH { GHC.Base.I# x_alG ->
238 GHC.Base.I# (GHC.Prim.+# x_alG 1)
240 T.$wfoo :: GHC.Prim.Int# -> (GHC.Base.Int -> GHC.Base.Int) ->
243 \ (ww_sme :: GHC.Prim.Int#) (w_smg :: GHC.Base.Int -> GHC.Base.Int) ->
244 case ww_sme of ds_Xlw {
246 case w_smg (GHC.Base.I# ds_Xlw) of w1_Xmo { GHC.Base.I# ww1_Xmz ->
247 T.$wfoo ww1_Xmz lvl_rmV
252 The recursive call has lvl_rmV as its argument, so we could create a specialised copy
253 with that argument baked in; that is, not passed at all. Now it can perhaps be inlined.
255 When is this worth it? Call the constant 'lvl'
256 - If 'lvl' has an unfolding that is a constructor, see if the corresponding
257 parameter is scrutinised anywhere in the body.
259 - If 'lvl' has an unfolding that is a inlinable function, see if the corresponding
260 parameter is applied (...to enough arguments...?)
262 Also do this is if the function has RULES?
266 Note [Specialising for lambda parameters]
267 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
268 foo :: Int -> (Int -> Int) -> Int
270 foo m f = foo (f m) (\n -> n-m)
272 This is subtly different from the previous one in that we get an
273 explicit lambda as the argument:
275 T.$wfoo :: GHC.Prim.Int# -> (GHC.Base.Int -> GHC.Base.Int) ->
278 \ (ww_sm8 :: GHC.Prim.Int#) (w_sma :: GHC.Base.Int -> GHC.Base.Int) ->
279 case ww_sm8 of ds_Xlr {
281 case w_sma (GHC.Base.I# ds_Xlr) of w1_Xmf { GHC.Base.I# ww1_Xmq ->
284 (\ (n_ad3 :: GHC.Base.Int) ->
285 case n_ad3 of wild_alB { GHC.Base.I# x_alA ->
286 GHC.Base.I# (GHC.Prim.-# x_alA ds_Xlr)
292 I wonder if SpecConstr couldn't be extended to handle this? After all,
293 lambda is a sort of constructor for functions and perhaps it already
294 has most of the necessary machinery?
296 Furthermore, there's an immediate win, because you don't need to allocate the lamda
297 at the call site; and if perchance it's called in the recursive call, then you
298 may avoid allocating it altogether. Just like for constructors.
300 Looks cool, but probably rare...but it might be easy to implement.
302 -----------------------------------------------------
303 Stuff not yet handled
304 -----------------------------------------------------
306 Here are notes arising from Roman's work that I don't want to lose.
312 foo :: Int -> T Int -> Int
314 foo x t | even x = case t of { T n -> foo (x-n) t }
315 | otherwise = foo (x-1) t
317 SpecConstr does no specialisation, because the second recursive call
318 looks like a boxed use of the argument. A pity.
320 $wfoo_sFw :: GHC.Prim.Int# -> T.T GHC.Base.Int -> GHC.Prim.Int#
322 \ (ww_sFo [Just L] :: GHC.Prim.Int#) (w_sFq [Just L] :: T.T GHC.Base.Int) ->
323 case ww_sFo of ds_Xw6 [Just L] {
325 case GHC.Prim.remInt# ds_Xw6 2 of wild1_aEF [Dead Just A] {
326 __DEFAULT -> $wfoo_sFw (GHC.Prim.-# ds_Xw6 1) w_sFq;
328 case w_sFq of wild_Xy [Just L] { T.T n_ad5 [Just U(L)] ->
329 case n_ad5 of wild1_aET [Just A] { GHC.Base.I# y_aES [Just L] ->
330 $wfoo_sFw (GHC.Prim.-# ds_Xw6 y_aES) wild_Xy
336 data a :*: b = !a :*: !b
339 foo :: (Int :*: T Int) -> Int
341 foo (x :*: t) | even x = case t of { T n -> foo ((x-n) :*: t) }
342 | otherwise = foo ((x-1) :*: t)
344 Very similar to the previous one, except that the parameters are now in
345 a strict tuple. Before SpecConstr, we have
347 $wfoo_sG3 :: GHC.Prim.Int# -> T.T GHC.Base.Int -> GHC.Prim.Int#
349 \ (ww_sFU [Just L] :: GHC.Prim.Int#) (ww_sFW [Just L] :: T.T
351 case ww_sFU of ds_Xws [Just L] {
353 case GHC.Prim.remInt# ds_Xws 2 of wild1_aEZ [Dead Just A] {
355 case ww_sFW of tpl_B2 [Just L] { T.T a_sFo [Just A] ->
356 $wfoo_sG3 (GHC.Prim.-# ds_Xws 1) tpl_B2 -- $wfoo1
359 case ww_sFW of wild_XB [Just A] { T.T n_ad7 [Just S(L)] ->
360 case n_ad7 of wild1_aFd [Just L] { GHC.Base.I# y_aFc [Just L] ->
361 $wfoo_sG3 (GHC.Prim.-# ds_Xws y_aFc) wild_XB -- $wfoo2
365 We get two specialisations:
366 "SC:$wfoo1" [0] __forall {a_sFB :: GHC.Base.Int sc_sGC :: GHC.Prim.Int#}
367 Foo.$wfoo sc_sGC (Foo.T @ GHC.Base.Int a_sFB)
368 = Foo.$s$wfoo1 a_sFB sc_sGC ;
369 "SC:$wfoo2" [0] __forall {y_aFp :: GHC.Prim.Int# sc_sGC :: GHC.Prim.Int#}
370 Foo.$wfoo sc_sGC (Foo.T @ GHC.Base.Int (GHC.Base.I# y_aFp))
371 = Foo.$s$wfoo y_aFp sc_sGC ;
373 But perhaps the first one isn't good. After all, we know that tpl_B2 is
374 a T (I# x) really, because T is strict and Int has one constructor. (We can't
375 unbox the strict fields, becuase T is polymorphic!)
379 %************************************************************************
381 \subsection{Top level wrapper stuff}
383 %************************************************************************
386 specConstrProgram :: DynFlags -> UniqSupply -> [CoreBind] -> IO [CoreBind]
387 specConstrProgram dflags us binds
389 showPass dflags "SpecConstr"
391 let (binds', _) = initUs us (go emptyScEnv binds)
393 endPass dflags "SpecConstr" Opt_D_dump_spec binds'
395 dumpIfSet_dyn dflags Opt_D_dump_rules "Top-level specialisations"
396 (pprRules (tidyRules emptyTidyEnv (rulesOfBinds binds')))
400 go env [] = returnUs []
401 go env (bind:binds) = scBind env bind `thenUs` \ (env', _, bind') ->
402 go env' binds `thenUs` \ binds' ->
403 returnUs (bind' : binds')
407 %************************************************************************
409 \subsection{Environment: goes downwards}
411 %************************************************************************
414 data ScEnv = SCE { scope :: InScopeEnv,
415 -- Binds all non-top-level variables in scope
420 type InScopeEnv = VarEnv HowBound
422 type ConstrEnv = IdEnv ConValue
423 data ConValue = CV AltCon [CoreArg]
424 -- Variables known to be bound to a constructor
425 -- in a particular case alternative
428 instance Outputable ConValue where
429 ppr (CV con args) = ppr con <+> interpp'SP args
431 refineConstrEnv :: Subst -> ConstrEnv -> ConstrEnv
432 -- The substitution is a type substitution only
433 refineConstrEnv subst env = mapVarEnv refine_con_value env
435 refine_con_value (CV con args) = CV con (map (substExpr subst) args)
437 emptyScEnv = SCE { scope = emptyVarEnv, cons = emptyVarEnv }
439 data HowBound = RecFun -- These are the recursive functions for which
440 -- we seek interesting call patterns
442 | RecArg -- These are those functions' arguments, or their sub-components;
443 -- we gather occurrence information for these
445 | Other -- We track all others so we know what's in scope
446 -- This is used in spec_one to check what needs to be
447 -- passed as a parameter and what is in scope at the
448 -- function definition site
450 instance Outputable HowBound where
451 ppr RecFun = text "RecFun"
452 ppr RecArg = text "RecArg"
453 ppr Other = text "Other"
455 lookupScopeEnv env v = lookupVarEnv (scope env) v
457 extendBndrs env bndrs = env { scope = extendVarEnvList (scope env) [(b,Other) | b <- bndrs] }
458 extendBndr env bndr = env { scope = extendVarEnv (scope env) bndr Other }
463 -- we want to bind b, and perhaps scrut too, to (C x y)
464 extendCaseBndrs :: ScEnv -> Id -> CoreExpr -> AltCon -> [Var] -> ScEnv
465 extendCaseBndrs env case_bndr scrut con alt_bndrs
468 LitAlt lit -> extendCons env1 scrut case_bndr (CV con [])
469 DataAlt dc -> extend_data_con dc
471 cur_scope = scope env
472 env1 = env { scope = extendVarEnvList cur_scope
473 [(b,how_bound) | b <- case_bndr:alt_bndrs] }
475 -- Record RecArg for the components iff the scrutinee is RecArg
476 -- [This comment looks plain wrong to me, so I'm ignoring it
477 -- "Also forget if the scrutinee is a RecArg, because we're
478 -- now in the branch of a case, and we don't want to
479 -- record a non-scrutinee use of v if we have
480 -- case v of { (a,b) -> ...(f v)... }" ]
481 how_bound = case scrut of
482 Var v -> lookupVarEnv cur_scope v `orElse` Other
485 extend_data_con data_con =
486 extendCons env1 scrut case_bndr (CV con vanilla_args)
488 vanilla_args = map Type (tyConAppArgs (idType case_bndr)) ++
489 varsToCoreExprs alt_bndrs
491 extendCons :: ScEnv -> CoreExpr -> Id -> ConValue -> ScEnv
492 extendCons env scrut case_bndr val
494 Var v -> env { cons = extendVarEnv cons1 v val }
495 other -> env { cons = cons1 }
497 cons1 = extendVarEnv (cons env) case_bndr val
499 -- When we encounter a recursive function binding
501 -- we want to extend the scope env with bindings
502 -- that record that f is a RecFn and x,y are RecArgs
503 extendRecBndr env fn bndrs
504 = env { scope = scope env `extendVarEnvList`
505 ((fn,RecFun): [(bndr,RecArg) | bndr <- bndrs]) }
509 %************************************************************************
511 \subsection{Usage information: flows upwards}
513 %************************************************************************
518 calls :: !(IdEnv ([Call])), -- Calls
519 -- The functions are a subset of the
520 -- RecFuns in the ScEnv
522 occs :: !(IdEnv ArgOcc) -- Information on argument occurrences
523 } -- The variables are a subset of the
524 -- RecArg in the ScEnv
526 type Call = (ConstrEnv, [CoreArg])
527 -- The arguments of the call, together with the
528 -- env giving the constructor bindings at the call site
530 nullUsage = SCU { calls = emptyVarEnv, occs = emptyVarEnv }
532 combineUsage u1 u2 = SCU { calls = plusVarEnv_C (++) (calls u1) (calls u2),
533 occs = plusVarEnv_C combineOcc (occs u1) (occs u2) }
535 combineUsages [] = nullUsage
536 combineUsages us = foldr1 combineUsage us
538 lookupOcc :: ScUsage -> Var -> (ScUsage, ArgOcc)
539 lookupOcc (SCU { calls = sc_calls, occs = sc_occs }) bndr
540 = (SCU {calls = sc_calls, occs = delVarEnv sc_occs bndr},
541 lookupVarEnv sc_occs bndr `orElse` NoOcc)
543 lookupOccs :: ScUsage -> [Var] -> (ScUsage, [ArgOcc])
544 lookupOccs (SCU { calls = sc_calls, occs = sc_occs }) bndrs
545 = (SCU {calls = sc_calls, occs = delVarEnvList sc_occs bndrs},
546 [lookupVarEnv sc_occs b `orElse` NoOcc | b <- bndrs])
548 data ArgOcc = NoOcc -- Doesn't occur at all; or a type argument
549 | UnkOcc -- Used in some unknown way
551 | ScrutOcc (UniqFM [ArgOcc]) -- See Note [ScrutOcc]
553 | BothOcc -- Definitely taken apart, *and* perhaps used in some other way
557 An occurrence of ScrutOcc indicates that the thing is *only* taken apart or applied.
559 Functions, litersl: ScrutOcc emptyUFM
560 Data constructors: ScrutOcc subs,
562 where (subs :: UniqFM [ArgOcc]) gives usage of the *pattern-bound* components,
563 The domain of the UniqFM is the Unique of the data constructor
565 The [ArgOcc] is the occurrences of the *pattern-bound* components
566 of the data structure. E.g.
567 data T a = forall b. MkT a b (b->a)
568 A pattern binds b, x::a, y::b, z::b->a, but not 'a'!
572 instance Outputable ArgOcc where
573 ppr (ScrutOcc xs) = ptext SLIT("scrut-occ") <> parens (ppr xs)
574 ppr UnkOcc = ptext SLIT("unk-occ")
575 ppr BothOcc = ptext SLIT("both-occ")
576 ppr NoOcc = ptext SLIT("no-occ")
578 combineOcc NoOcc occ = occ
579 combineOcc occ NoOcc = occ
580 combineOcc (ScrutOcc xs) (ScrutOcc ys) = ScrutOcc (plusUFM_C combineOccs xs ys)
581 combineOcc UnkOcc UnkOcc = UnkOcc
582 combineOcc _ _ = BothOcc
584 combineOccs :: [ArgOcc] -> [ArgOcc] -> [ArgOcc]
585 combineOccs xs ys = zipWithEqual "combineOccs" combineOcc xs ys
587 conArgOccs :: ArgOcc -> AltCon -> [ArgOcc]
588 -- Find usage of components of data con; returns [UnkOcc...] if unknown
589 -- See Note [ScrutOcc] for the extra UnkOccs in the vanilla datacon case
591 conArgOccs (ScrutOcc fm) (DataAlt dc)
592 | Just pat_arg_occs <- lookupUFM fm dc
593 = tyvar_unks ++ pat_arg_occs
595 tyvar_unks | isVanillaDataCon dc = [UnkOcc | tv <- dataConTyVars dc]
598 conArgOccs other con = repeat UnkOcc
602 %************************************************************************
604 \subsection{The main recursive function}
606 %************************************************************************
608 The main recursive function gathers up usage information, and
609 creates specialised versions of functions.
612 scExpr :: ScEnv -> CoreExpr -> UniqSM (ScUsage, CoreExpr)
613 -- The unique supply is needed when we invent
614 -- a new name for the specialised function and its args
616 scExpr env e@(Type t) = returnUs (nullUsage, e)
617 scExpr env e@(Lit l) = returnUs (nullUsage, e)
618 scExpr env e@(Var v) = returnUs (varUsage env v UnkOcc, e)
619 scExpr env (Note n e) = scExpr env e `thenUs` \ (usg,e') ->
620 returnUs (usg, Note n e')
621 scExpr env (Cast e co)= scExpr env e `thenUs` \ (usg,e') ->
622 returnUs (usg, Cast e' co)
623 scExpr env (Lam b e) = scExpr (extendBndr env b) e `thenUs` \ (usg,e') ->
624 returnUs (usg, Lam b e')
626 scExpr env (Case scrut b ty alts)
627 = do { (alt_usgs, alt_occs, alts') <- mapAndUnzip3Us sc_alt alts
628 ; let (alt_usg, b_occ) = lookupOcc (combineUsages alt_usgs) b
629 scrut_occ = foldr combineOcc b_occ alt_occs
630 -- The combined usage of the scrutinee is given
631 -- by scrut_occ, which is passed to scScrut, which
632 -- in turn treats a bare-variable scrutinee specially
633 ; (scrut_usg, scrut') <- scScrut env scrut scrut_occ
634 ; return (alt_usg `combineUsage` scrut_usg,
635 Case scrut' b ty alts') }
638 = do { let env1 = extendCaseBndrs env b scrut con bs
639 ; (usg,rhs') <- scExpr env1 rhs
640 ; let (usg', arg_occs) = lookupOccs usg bs
641 scrut_occ = case con of
642 DataAlt dc -> ScrutOcc (unitUFM dc arg_occs)
643 other -> ScrutOcc emptyUFM
644 ; return (usg', scrut_occ, (con,bs,rhs')) }
646 scExpr env (Let bind body)
647 = scBind env bind `thenUs` \ (env', bind_usg, bind') ->
648 scExpr env' body `thenUs` \ (body_usg, body') ->
649 returnUs (bind_usg `combineUsage` body_usg, Let bind' body')
651 scExpr env e@(App _ _)
652 = do { let (fn, args) = collectArgs e
653 ; (fn_usg, fn') <- scScrut env fn (ScrutOcc emptyUFM)
654 -- Process the function too. It's almost always a variable,
655 -- but not always. In particular, if this pass follows float-in,
656 -- which it may, we can get
657 -- (let f = ...f... in f) arg1 arg2
658 -- We use scScrut to record the fact that the function is called
659 -- Perhpas we should check that it has at least one value arg,
660 -- but currently we don't bother
662 ; (arg_usgs, args') <- mapAndUnzipUs (scExpr env) args
663 ; let call_usg = case fn of
664 Var f | Just RecFun <- lookupScopeEnv env f
665 -> SCU { calls = unitVarEnv f [(cons env, args)],
668 ; return (combineUsages arg_usgs `combineUsage` fn_usg
669 `combineUsage` call_usg,
673 ----------------------
674 scScrut :: ScEnv -> CoreExpr -> ArgOcc -> UniqSM (ScUsage, CoreExpr)
675 -- Used for the scrutinee of a case,
676 -- or the function of an application
677 scScrut env e@(Var v) occ = returnUs (varUsage env v occ, e)
678 scScrut env e occ = scExpr env e
681 ----------------------
682 scBind :: ScEnv -> CoreBind -> UniqSM (ScEnv, ScUsage, CoreBind)
683 scBind env (Rec [(fn,rhs)])
685 = scExpr env_fn_body body `thenUs` \ (usg, body') ->
686 specialise env fn bndrs body' usg `thenUs` \ (rules, spec_prs) ->
687 -- Note body': the specialised copies should be based on the
688 -- optimised version of the body, in case there were
689 -- nested functions inside.
691 SCU { calls = calls, occs = occs } = usg
693 returnUs (extendBndr env fn, -- For the body of the letrec, just
694 -- extend the env with Other to record
695 -- that it's in scope; no funny RecFun business
696 SCU { calls = calls `delVarEnv` fn, occs = occs `delVarEnvList` val_bndrs},
697 Rec ((fn `addIdSpecialisations` rules, mkLams bndrs body') : spec_prs))
699 (bndrs,body) = collectBinders rhs
700 val_bndrs = filter isId bndrs
701 env_fn_body = extendRecBndr env fn bndrs
704 = mapAndUnzipUs do_one prs `thenUs` \ (usgs, prs') ->
705 returnUs (extendBndrs env (map fst prs), combineUsages usgs, Rec prs')
707 do_one (bndr,rhs) = scExpr env rhs `thenUs` \ (usg, rhs') ->
708 returnUs (usg, (bndr,rhs'))
710 scBind env (NonRec bndr rhs)
711 = scExpr env rhs `thenUs` \ (usg, rhs') ->
712 returnUs (extendBndr env bndr, usg, NonRec bndr rhs')
714 ----------------------
716 | Just RecArg <- lookupScopeEnv env v = SCU { calls = emptyVarEnv,
717 occs = unitVarEnv v use }
718 | otherwise = nullUsage
722 %************************************************************************
724 \subsection{The specialiser}
726 %************************************************************************
731 -> [CoreBndr] -> CoreExpr -- Its RHS
732 -> ScUsage -- Info on usage
733 -> UniqSM ([CoreRule], -- Rules
734 [(Id,CoreExpr)]) -- Bindings
736 specialise env fn bndrs body body_usg
737 = do { let (_, bndr_occs) = lookupOccs body_usg bndrs
739 ; mb_calls <- mapM (callToPats (scope env) bndr_occs)
740 (lookupVarEnv (calls body_usg) fn `orElse` [])
742 ; let good_calls :: [([Var], [CoreArg])]
743 good_calls = catMaybes mb_calls
744 in_scope = mkInScopeSet $ unionVarSets $
745 [ exprsFreeVars pats `delVarSetList` vs
746 | (vs,pats) <- good_calls ]
747 uniq_calls = nubBy (same_call in_scope) good_calls
749 mapAndUnzipUs (spec_one env fn (mkLams bndrs body))
750 (uniq_calls `zip` [1..]) }
752 -- Two calls are the same if they match both ways
753 same_call in_scope (vs1,as1)(vs2,as2)
754 = isJust (matchN in_scope vs1 as1 as2)
755 && isJust (matchN in_scope vs2 as2 as1)
757 callToPats :: InScopeEnv -> [ArgOcc] -> Call
758 -> UniqSM (Maybe ([Var], [CoreExpr]))
759 -- The VarSet is the variables to quantify over in the rule
760 -- The [CoreExpr] are the argument patterns for the rule
761 callToPats in_scope bndr_occs (con_env, args)
762 | length args < length bndr_occs -- Check saturated
765 = do { prs <- argsToPats in_scope con_env (args `zip` bndr_occs)
766 ; let (good_pats, pats) = unzip prs
767 pat_fvs = varSetElems (exprsFreeVars pats)
768 qvars = filter (not . (`elemVarEnv` in_scope)) pat_fvs
769 -- Quantify over variables that are not in sccpe
770 -- See Note [Shadowing] at the top
773 then return (Just (qvars, pats))
774 else return Nothing }
776 ---------------------
779 -> CoreExpr -- Rhs of the original function
780 -> (([Var], [CoreArg]), Int)
781 -> UniqSM (CoreRule, (Id,CoreExpr)) -- Rule and binding
783 -- spec_one creates a specialised copy of the function, together
784 -- with a rule for using it. I'm very proud of how short this
785 -- function is, considering what it does :-).
791 f = /\b \y::[(a,b)] -> ....f (b,c) ((:) (a,(b,c)) (x,v) (h w))...
792 [c::*, v::(b,c) are presumably bound by the (...) part]
794 f_spec = /\ b c \ v::(b,c) hw::[(a,(b,c))] ->
795 (...entire RHS of f...) (b,c) ((:) (a,(b,c)) (x,v) hw)
797 RULE: forall b::* c::*, -- Note, *not* forall a, x
801 f (b,c) ((:) (a,(b,c)) (x,v) hw) = f_spec b c v hw
804 spec_one env fn rhs ((vars_to_bind, pats), rule_number)
805 = getUniqueUs `thenUs` \ spec_uniq ->
808 fn_loc = nameSrcLoc fn_name
809 spec_occ = mkSpecOcc (nameOccName fn_name)
811 -- Put the type variables first; the type of a term
812 -- variable may mention a type variable
813 (tvs, ids) = partition isTyVar vars_to_bind
815 spec_body = mkApps rhs pats
816 body_ty = exprType spec_body
818 (spec_lam_args, spec_call_args) = mkWorkerArgs bndrs body_ty
819 -- Usual w/w hack to avoid generating
820 -- a spec_rhs of unlifted type and no args
822 rule_name = mkFastString ("SC:" ++ showSDoc (ppr fn <> int rule_number))
823 spec_rhs = mkLams spec_lam_args spec_body
824 spec_id = mkUserLocal spec_occ spec_uniq (mkPiTypes spec_lam_args body_ty) fn_loc
825 rule_rhs = mkVarApps (Var spec_id) spec_call_args
826 rule = mkLocalRule rule_name specConstrActivation fn_name bndrs pats rule_rhs
828 returnUs (rule, (spec_id, spec_rhs))
830 -- In which phase should the specialise-constructor rules be active?
831 -- Originally I made them always-active, but Manuel found that
832 -- this defeated some clever user-written rules. So Plan B
833 -- is to make them active only in Phase 0; after all, currently,
834 -- the specConstr transformation is only run after the simplifier
835 -- has reached Phase 0. In general one would want it to be
836 -- flag-controllable, but for now I'm leaving it baked in
838 specConstrActivation :: Activation
839 specConstrActivation = ActiveAfter 0 -- Baked in; see comments above
842 %************************************************************************
844 \subsection{Argument analysis}
846 %************************************************************************
848 This code deals with analysing call-site arguments to see whether
849 they are constructor applications.
853 -- argToPat takes an actual argument, and returns an abstracted
854 -- version, consisting of just the "constructor skeleton" of the
855 -- argument, with non-constructor sub-expression replaced by new
856 -- placeholder variables. For example:
857 -- C a (D (f x) (g y)) ==> C p1 (D p2 p3)
859 argToPat :: InScopeEnv -- What's in scope at the fn defn site
860 -> ConstrEnv -- ConstrEnv at the call site
861 -> CoreArg -- A call arg (or component thereof)
863 -> UniqSM (Bool, CoreArg)
864 -- Returns (interesting, pat),
865 -- where pat is the pattern derived from the argument
866 -- intersting=True if the pattern is non-trivial (not a variable or type)
867 -- E.g. x:xs --> (True, x:xs)
868 -- f xs --> (False, w) where w is a fresh wildcard
869 -- (f xs, 'c') --> (True, (w, 'c')) where w is a fresh wildcard
870 -- \x. x+y --> (True, \x. x+y)
871 -- lvl7 --> (True, lvl7) if lvl7 is bound
872 -- somewhere further out
874 argToPat in_scope con_env arg@(Type ty) arg_occ
875 = return (False, arg)
877 argToPat in_scope con_env (Var v) arg_occ
878 | not (isLocalId v) || v `elemVarEnv` in_scope
879 = -- The recursive call passes a variable that
880 -- is in scope at the function definition site
881 -- It's worth specialising on this if
882 -- (a) it's used in an interesting way in the body
883 -- (b) we know what its value is
884 if (case arg_occ of { UnkOcc -> False; other -> True }) -- (a)
885 && isValueUnfolding (idUnfolding v) -- (b)
886 then return (True, Var v)
887 else wildCardPat (idType v)
889 argToPat in_scope con_env arg arg_occ
893 is_value_lam (Lam v e) -- Spot a value lambda, even if
894 | isId v = True -- it is inside a type lambda
895 | otherwise = is_value_lam e
896 is_value_lam other = False
898 argToPat in_scope con_env arg arg_occ
899 | Just (CV dc args) <- is_con_app_maybe con_env arg
901 ScrutOcc _ -> True -- Used only by case scrutinee
902 BothOcc -> case arg of -- Used by case scrut
903 App {} -> True -- ...and elsewhere...
905 other -> False -- No point; the arg is not decomposed
906 = do { args' <- argsToPats in_scope con_env (args `zip` conArgOccs arg_occ dc)
907 ; return (True, mk_con_app dc (map snd args')) }
909 argToPat in_scope con_env (Var v) arg_occ
910 = -- A variable bound inside the function.
911 -- Don't make a wild-card, because we may usefully share
912 -- e.g. f a = let x = ... in f (x,x)
913 -- NB: this case follows the lambda and con-app cases!!
914 return (False, Var v)
916 -- The default case: make a wild-card
917 argToPat in_scope con_env arg arg_occ = wildCardPat (exprType arg)
919 wildCardPat :: Type -> UniqSM (Bool, CoreArg)
920 wildCardPat ty = do { uniq <- getUniqueUs
921 ; let id = mkSysLocal FSLIT("sc") uniq ty
922 ; return (False, Var id) }
924 argsToPats :: InScopeEnv -> ConstrEnv
925 -> [(CoreArg, ArgOcc)]
926 -> UniqSM [(Bool, CoreArg)]
927 argsToPats in_scope con_env args
930 do_one (arg,occ) = argToPat in_scope con_env arg occ
935 is_con_app_maybe :: ConstrEnv -> CoreExpr -> Maybe ConValue
936 is_con_app_maybe env (Var v)
937 = case lookupVarEnv env v of
938 Just stuff -> Just stuff
939 -- You might think we could look in the idUnfolding here
940 -- but that doesn't take account of which branch of a
941 -- case we are in, which is the whole point
943 Nothing | isCheapUnfolding unf
944 -> is_con_app_maybe env (unfoldingTemplate unf)
947 -- However we do want to consult the unfolding
948 -- as well, for let-bound constructors!
952 is_con_app_maybe env (Lit lit)
953 = Just (CV (LitAlt lit) [])
955 is_con_app_maybe env expr
956 = case collectArgs expr of
957 (Var fun, args) | Just con <- isDataConWorkId_maybe fun,
958 args `lengthAtLeast` dataConRepArity con
959 -- Might be > because the arity excludes type args
960 -> Just (CV (DataAlt con) args)
964 mk_con_app :: AltCon -> [CoreArg] -> CoreExpr
965 mk_con_app (LitAlt lit) [] = Lit lit
966 mk_con_app (DataAlt con) args = mkConApp con args