2 % (c) The GRASP/AQUA Project, Glasgow University, 1992-1998
4 \section{Tidying up Core}
7 module TidyPgm( mkBootModDetailsDs, mkBootModDetailsTc,
8 tidyProgram, globaliseAndTidyId ) where
10 #include "HsVersions.h"
22 import CoreArity ( exprArity )
23 import Class ( classSelIds )
45 import FastBool hiding ( fastOr )
48 import Data.List ( sortBy )
49 import Data.IORef ( IORef, readIORef, writeIORef )
53 Constructing the TypeEnv, Instances, Rules from which the ModIface is
54 constructed, and which goes on to subsequent modules in --make mode.
56 Most of the interface file is obtained simply by serialising the
57 TypeEnv. One important consequence is that if the *interface file*
58 has pragma info if and only if the final TypeEnv does. This is not so
59 important for *this* module, but it's essential for ghc --make:
60 subsequent compilations must not see (e.g.) the arity if the interface
61 file does not contain arity If they do, they'll exploit the arity;
62 then the arity might change, but the iface file doesn't change =>
63 recompilation does not happen => disaster.
65 For data types, the final TypeEnv will have a TyThing for the TyCon,
66 plus one for each DataCon; the interface file will contain just one
67 data type declaration, but it is de-serialised back into a collection
70 %************************************************************************
74 %************************************************************************
77 Plan A: mkBootModDetails: omit pragmas, make interfaces small
78 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
81 * Drop all WiredIn things from the TypeEnv
82 (we never want them in interface files)
84 * Retain all TyCons and Classes in the TypeEnv, to avoid
85 having to find which ones are mentioned in the
88 * Trim off the constructors of non-exported TyCons, both
89 from the TyCon and from the TypeEnv
91 * Drop non-exported Ids from the TypeEnv
93 * Tidy the types of the DFunIds of Instances,
94 make them into GlobalIds, (they already have External Names)
95 and add them to the TypeEnv
97 * Tidy the types of the (exported) Ids in the TypeEnv,
98 make them into GlobalIds (they already have External Names)
100 * Drop rules altogether
102 * Tidy the bindings, to ensure that the Caf and Arity
103 information is correct for each top-level binder; the
104 code generator needs it. And to ensure that local names have
105 distinct OccNames in case of object-file splitting
108 -- This is Plan A: make a small type env when typechecking only,
109 -- or when compiling a hs-boot file, or simply when not using -O
111 -- We don't look at the bindings at all -- there aren't any
114 mkBootModDetailsTc :: HscEnv -> TcGblEnv -> IO ModDetails
115 mkBootModDetailsTc hsc_env
116 TcGblEnv{ tcg_exports = exports,
117 tcg_type_env = type_env,
119 tcg_fam_insts = fam_insts
121 = mkBootModDetails hsc_env exports type_env insts fam_insts
123 mkBootModDetailsDs :: HscEnv -> ModGuts -> IO ModDetails
124 mkBootModDetailsDs hsc_env
125 ModGuts{ mg_exports = exports,
128 mg_fam_insts = fam_insts
130 = mkBootModDetails hsc_env exports type_env insts fam_insts
132 mkBootModDetails :: HscEnv -> [AvailInfo] -> NameEnv TyThing
133 -> [Instance] -> [FamInstEnv.FamInst] -> IO ModDetails
134 mkBootModDetails hsc_env exports type_env insts fam_insts
135 = do { let dflags = hsc_dflags hsc_env
136 ; showPass dflags "Tidy [hoot] type env"
138 ; let { insts' = tidyInstances globaliseAndTidyId insts
139 ; dfun_ids = map instanceDFunId insts'
140 ; type_env1 = tidyBootTypeEnv (availsToNameSet exports) type_env
141 ; type_env' = extendTypeEnvWithIds type_env1 dfun_ids
143 ; return (ModDetails { md_types = type_env'
145 , md_fam_insts = fam_insts
148 , md_exports = exports
149 , md_vect_info = noVectInfo
154 tidyBootTypeEnv :: NameSet -> TypeEnv -> TypeEnv
155 tidyBootTypeEnv exports type_env
156 = tidyTypeEnv True False exports type_env final_ids
158 -- Find the LocalIds in the type env that are exported
159 -- Make them into GlobalIds, and tidy their types
161 -- It's very important to remove the non-exported ones
162 -- because we don't tidy the OccNames, and if we don't remove
163 -- the non-exported ones we'll get many things with the
164 -- same name in the interface file, giving chaos.
165 final_ids = [ globaliseAndTidyId id
166 | id <- typeEnvIds type_env
170 -- default methods have their export flag set, but everything
171 -- else doesn't (yet), because this is pre-desugaring, so we
173 keep_it id = isExportedId id || idName id `elemNameSet` exports
177 globaliseAndTidyId :: Id -> Id
178 -- Takes an LocalId with an External Name,
179 -- makes it into a GlobalId
180 -- * unchanged Name (might be Internal or External)
181 -- * unchanged details
182 -- * VanillaIdInfo (makes a conservative assumption about Caf-hood)
183 globaliseAndTidyId id
184 = Id.setIdType (globaliseId id) tidy_type
186 tidy_type = tidyTopType (idType id)
190 %************************************************************************
192 Plan B: tidy bindings, make TypeEnv full of IdInfo
194 %************************************************************************
196 Plan B: include pragmas, make interfaces
197 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
198 * Figure out which Ids are externally visible
200 * Tidy the bindings, externalising appropriate Ids
202 * Drop all Ids from the TypeEnv, and add all the External Ids from
203 the bindings. (This adds their IdInfo to the TypeEnv; and adds
204 floated-out Ids that weren't even in the TypeEnv before.)
206 Step 1: Figure out external Ids
207 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
208 Note [choosing external names]
210 First we figure out which Ids are "external" Ids. An
211 "external" Id is one that is visible from outside the compilation
213 a) the user exported ones
214 b) ones mentioned in the unfoldings, workers,
215 or rules of externally-visible ones
217 While figuring out which Ids are external, we pick a "tidy" OccName
218 for each one. That is, we make its OccName distinct from the other
219 external OccNames in this module, so that in interface files and
220 object code we can refer to it unambiguously by its OccName. The
221 OccName for each binder is prefixed by the name of the exported Id
222 that references it; e.g. if "f" references "x" in its unfolding, then
223 "x" is renamed to "f_x". This helps distinguish the different "x"s
224 from each other, and means that if "f" is later removed, things that
225 depend on the other "x"s will not need to be recompiled. Of course,
226 if there are multiple "f_x"s, then we have to disambiguate somehow; we
227 use "f_x0", "f_x1" etc.
229 As far as possible we should assign names in a deterministic fashion.
230 Each time this module is compiled with the same options, we should end
231 up with the same set of external names with the same types. That is,
232 the ABI hash in the interface should not change. This turns out to be
233 quite tricky, since the order of the bindings going into the tidy
234 phase is already non-deterministic, as it is based on the ordering of
235 Uniques, which are assigned unpredictably.
237 To name things in a stable way, we do a depth-first-search of the
238 bindings, starting from the exports sorted by name. This way, as long
239 as the bindings themselves are deterministic (they sometimes aren't!),
240 the order in which they are presented to the tidying phase does not
241 affect the names we assign.
243 Step 2: Tidy the program
244 ~~~~~~~~~~~~~~~~~~~~~~~~
245 Next we traverse the bindings top to bottom. For each *top-level*
248 1. Make it into a GlobalId; its IdDetails becomes VanillaGlobal,
249 reflecting the fact that from now on we regard it as a global,
252 2. Give it a system-wide Unique.
253 [Even non-exported things need system-wide Uniques because the
254 byte-code generator builds a single Name->BCO symbol table.]
256 We use the NameCache kept in the HscEnv as the
257 source of such system-wide uniques.
259 For external Ids, use the original-name cache in the NameCache
260 to ensure that the unique assigned is the same as the Id had
261 in any previous compilation run.
263 3. Rename top-level Ids according to the names we chose in step 1.
264 If it's an external Id, make it have a External Name, otherwise
265 make it have an Internal Name. This is used by the code generator
266 to decide whether to make the label externally visible
268 4. Give it its UTTERLY FINAL IdInfo; in ptic,
269 * its unfolding, if it should have one
271 * its arity, computed from the number of visible lambdas
273 * its CAF info, computed from what is free in its RHS
276 Finally, substitute these new top-level binders consistently
277 throughout, including in unfoldings. We also tidy binders in
278 RHSs, so that they print nicely in interfaces.
281 tidyProgram :: HscEnv -> ModGuts -> IO (CgGuts, ModDetails)
282 tidyProgram hsc_env (ModGuts { mg_module = mod, mg_exports = exports,
284 mg_insts = insts, mg_fam_insts = fam_insts,
286 mg_rules = imp_rules,
287 mg_vect_info = vect_info,
288 mg_dir_imps = dir_imps,
291 mg_foreign = foreign_stubs,
292 mg_hpc_info = hpc_info,
293 mg_modBreaks = modBreaks })
295 = do { let dflags = hsc_dflags hsc_env
296 ; showPass dflags "Tidy Core"
298 ; let { omit_prags = dopt Opt_OmitInterfacePragmas dflags
299 ; th = dopt Opt_TemplateHaskell dflags
302 ; (unfold_env, tidy_occ_env)
303 <- chooseExternalIds hsc_env type_env mod omit_prags binds
307 | otherwise = findExternalRules binds imp_rules unfold_env
308 -- findExternalRules filters imp_rules to avoid binders that
309 -- aren't externally visible; but the externally-visible binders
310 -- are computed (by findExternalIds) assuming that all orphan
311 -- rules are exported (they get their Exported flag set in the desugarer)
312 -- So in fact we may export more than we need.
313 -- (It's a sort of mutual recursion.)
316 ; let { (tidy_env, tidy_binds)
317 = tidyTopBinds hsc_env unfold_env tidy_occ_env binds }
319 ; let { export_set = availsToNameSet exports
320 ; final_ids = [ id | id <- bindersOfBinds tidy_binds,
321 isExternalName (idName id)]
322 ; tidy_type_env = tidyTypeEnv omit_prags th export_set
324 ; tidy_insts = tidyInstances (lookup_dfun tidy_type_env) insts
325 -- A DFunId will have a binding in tidy_binds, and so
326 -- will now be in final_env, replete with IdInfo
327 -- Its name will be unchanged since it was born, but
328 -- we want Global, IdInfo-rich (or not) DFunId in the
331 ; tidy_rules = tidyRules tidy_env ext_rules
332 -- You might worry that the tidy_env contains IdInfo-rich stuff
333 -- and indeed it does, but if omit_prags is on, ext_rules is
336 -- See Note [Injecting implicit bindings]
337 ; implicit_binds = getImplicitBinds type_env
338 ; all_tidy_binds = implicit_binds ++ tidy_binds
340 ; alg_tycons = filter isAlgTyCon (typeEnvTyCons type_env)
343 ; endPass dflags "Tidy Core" Opt_D_dump_simpl all_tidy_binds
344 ; dumpIfSet_core dflags Opt_D_dump_simpl
346 (pprRules tidy_rules)
348 ; let dir_imp_mods = moduleEnvKeys dir_imps
350 ; return (CgGuts { cg_module = mod,
351 cg_tycons = alg_tycons,
352 cg_binds = all_tidy_binds,
353 cg_dir_imps = dir_imp_mods,
354 cg_foreign = foreign_stubs,
355 cg_dep_pkgs = dep_pkgs deps,
356 cg_hpc_info = hpc_info,
357 cg_modBreaks = modBreaks },
359 ModDetails { md_types = tidy_type_env,
360 md_rules = tidy_rules,
361 md_insts = tidy_insts,
362 md_fam_insts = fam_insts,
363 md_exports = exports,
364 md_anns = anns, -- are already tidy
365 md_vect_info = vect_info --
369 lookup_dfun :: TypeEnv -> Var -> Id
370 lookup_dfun type_env dfun_id
371 = case lookupTypeEnv type_env (idName dfun_id) of
372 Just (AnId dfun_id') -> dfun_id'
373 _other -> pprPanic "lookup_dfun" (ppr dfun_id)
375 --------------------------
376 tidyTypeEnv :: Bool -- Compiling without -O, so omit prags
377 -> Bool -- Template Haskell is on
378 -> NameSet -> TypeEnv -> [Id] -> TypeEnv
380 -- The competed type environment is gotten from
381 -- Dropping any wired-in things, and then
382 -- a) keeping the types and classes
383 -- b) removing all Ids,
384 -- c) adding Ids with correct IdInfo, including unfoldings,
385 -- gotten from the bindings
386 -- From (c) we keep only those Ids with External names;
387 -- the CoreTidy pass makes sure these are all and only
388 -- the externally-accessible ones
389 -- This truncates the type environment to include only the
390 -- exported Ids and things needed from them, which saves space
392 tidyTypeEnv omit_prags th exports type_env final_ids
393 = let type_env1 = filterNameEnv keep_it type_env
394 type_env2 = extendTypeEnvWithIds type_env1 final_ids
395 type_env3 | omit_prags = mapNameEnv (trimThing th exports) type_env2
396 | otherwise = type_env2
400 -- We keep GlobalIds, because they won't appear
401 -- in the bindings from which final_ids are derived!
402 -- (The bindings bind LocalIds.)
403 keep_it thing | isWiredInThing thing = False
404 keep_it (AnId id) = isGlobalId id -- Keep GlobalIds (e.g. class ops)
405 keep_it _other = True -- Keep all TyCons, DataCons, and Classes
407 --------------------------
408 isWiredInThing :: TyThing -> Bool
409 isWiredInThing thing = isWiredInName (getName thing)
411 --------------------------
412 trimThing :: Bool -> NameSet -> TyThing -> TyThing
413 -- Trim off inessentials, for boot files and no -O
414 trimThing th exports (ATyCon tc)
415 | not th && not (mustExposeTyCon exports tc)
416 = ATyCon (makeTyConAbstract tc) -- Note [Trimming and Template Haskell]
418 trimThing _th _exports (AnId id)
419 | not (isImplicitId id)
420 = AnId (id `setIdInfo` vanillaIdInfo)
422 trimThing _th _exports other_thing
426 {- Note [Trimming and Template Haskell]
427 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
428 Consider (Trac #2386) this
429 module M(T, makeOne) where
431 makeOne = [| Yay "Yep" |]
432 Notice that T is exported abstractly, but makeOne effectively exports it too!
433 A module that splices in $(makeOne) will then look for a declartion of Yay,
434 so it'd better be there. Hence, brutally but simply, we switch off type
435 constructor trimming if TH is enabled in this module. -}
438 mustExposeTyCon :: NameSet -- Exports
439 -> TyCon -- The tycon
440 -> Bool -- Can its rep be hidden?
441 -- We are compiling without -O, and thus trying to write as little as
442 -- possible into the interface file. But we must expose the details of
443 -- any data types whose constructors or fields are exported
444 mustExposeTyCon exports tc
445 | not (isAlgTyCon tc) -- Synonyms
447 | isEnumerationTyCon tc -- For an enumeration, exposing the constructors
448 = True -- won't lead to the need for further exposure
449 -- (This includes data types with no constructors.)
450 | isOpenTyCon tc -- Open type family
453 | otherwise -- Newtype, datatype
454 = any exported_con (tyConDataCons tc)
455 -- Expose rep if any datacon or field is exported
457 || (isNewTyCon tc && isFFITy (snd (newTyConRhs tc)))
458 -- Expose the rep for newtypes if the rep is an FFI type.
459 -- For a very annoying reason. 'Foreign import' is meant to
460 -- be able to look through newtypes transparently, but it
461 -- can only do that if it can "see" the newtype representation
463 exported_con con = any (`elemNameSet` exports)
464 (dataConName con : dataConFieldLabels con)
466 tidyInstances :: (DFunId -> DFunId) -> [Instance] -> [Instance]
467 tidyInstances tidy_dfun ispecs
470 tidy ispec = setInstanceDFunId ispec $
471 tidy_dfun (instanceDFunId ispec)
475 %************************************************************************
479 %************************************************************************
481 Note [Injecting implicit bindings]
482 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
483 We inject the implict bindings right at the end, in CoreTidy.
484 Some of these bindings, notably record selectors, are not
485 constructed in an optimised form. E.g. record selector for
486 data T = MkT { x :: {-# UNPACK #-} !Int }
487 Then the unfolding looks like
488 x = \t. case t of MkT x1 -> let x = I# x1 in x
489 This generates bad code unless it's first simplified a bit. That is
490 why CoreUnfold.mkImplicitUnfolding uses simleExprOpt to do a bit of
491 optimisation first. (Only matters when the selector is used curried;
492 eg map x ys.) See Trac #2070.
494 At one time I tried injecting the implicit bindings *early*, at the
495 beginning of SimplCore. But that gave rise to real difficulty,
496 becuase GlobalIds are supposed to have *fixed* IdInfo, but the
497 simplifier and other core-to-core passes mess with IdInfo all the
498 time. The straw that broke the camels back was when a class selector
499 got the wrong arity -- ie the simplifier gave it arity 2, whereas
500 importing modules were expecting it to have arity 1 (Trac #2844).
501 It's much safer just to inject them right at the end, after tidying.
503 Oh: two other reasons for injecting them late:
504 - If implicit Ids are already in the bindings when we start TidyPgm,
505 we'd have to be careful not to treat them as external Ids (in
506 the sense of findExternalIds); else the Ids mentioned in *their*
507 RHSs will be treated as external and you get an interface file
509 but nothing refererring to a18 (because the implicit Id is the
512 - More seriously, the tidied type-envt will include the implicit
513 Id replete with a18 in its unfolding; but we won't take account
514 of a18 when computing a fingerprint for the class; result chaos.
518 getImplicitBinds :: TypeEnv -> [CoreBind]
519 getImplicitBinds type_env
520 = map get_defn (concatMap implicit_ids (typeEnvElts type_env))
522 implicit_ids (ATyCon tc) = mapCatMaybes dataConWrapId_maybe (tyConDataCons tc)
523 implicit_ids (AClass cls) = classSelIds cls
526 get_defn :: Id -> CoreBind
527 get_defn id = NonRec id (unfoldingTemplate (idUnfolding id))
531 %************************************************************************
533 \subsection{Step 1: finding externals}
535 %************************************************************************
537 Sete Note [choosing external names].
540 type UnfoldEnv = IdEnv (Name{-new name-}, Bool {-show unfolding-})
542 chooseExternalIds :: HscEnv
547 -> IO (UnfoldEnv, TidyOccEnv)
548 -- maps top-level Ids to new, renamed, Ids.
549 -- If the new Id is external, it will be visible
550 -- in the interface file.
551 -- Bool => expose unfolding or not.
552 -- Step 1 from the notes above
554 chooseExternalIds hsc_env type_env mod omit_prags binds
556 (unfold_env1,occ_env1)
557 <- search (zip sorted_exports sorted_exports) emptyVarEnv init_occ_env
558 let internal_ids = filter (not . (`elemVarEnv` unfold_env1)) binders
559 tidy_internal internal_ids unfold_env1 occ_env1
561 nc_var = hsc_NC hsc_env
563 -- the exports, sorted by OccName. This is a deterministic list of
564 -- Ids (i.e. it's the same list every time this module is compiled),
565 -- in contrast to the bindings, which are ordered
566 -- non-deterministically.
568 -- This list will serve as a starting point for finding a
569 -- deterministic, tidy, renaming for all external Ids in this
571 sorted_exports = sortBy (compare `on` getOccName) $
572 filter isExportedId binders
574 binders = bindersOfBinds binds
576 bind_env :: IdEnv CoreExpr
577 bind_env = mkVarEnv (flattenBinds binds)
579 avoids = [getOccName name | bndr <- typeEnvIds type_env,
580 let name = idName bndr,
582 -- In computing our "avoids" list, we must include
584 -- all things with global names (assigned once and for
585 -- all by the renamer)
586 -- since their names are "taken".
587 -- The type environment is a convenient source of such things.
589 -- We also make sure to avoid any exported binders. Consider
590 -- f{-u1-} = 1 -- Local decl
592 -- f{-u2-} = 2 -- Exported decl
594 -- The second exported decl must 'get' the name 'f', so we
595 -- have to put 'f' in the avoids list before we get to the first
596 -- decl. tidyTopId then does a no-op on exported binders.
597 init_occ_env = initTidyOccEnv avoids
600 search :: [(Id,Id)] -- (external id, referrring id)
601 -> UnfoldEnv -- id -> (new Name, show_unfold)
602 -> TidyOccEnv -- occ env for choosing new Names
603 -> IO (UnfoldEnv, TidyOccEnv)
605 search [] unfold_env occ_env = return (unfold_env, occ_env)
607 search ((id,referrer) : rest) unfold_env occ_env
608 | id `elemVarEnv` unfold_env = search rest unfold_env occ_env
610 (occ_env', name') <- tidyTopName mod nc_var (Just referrer) occ_env id
612 rhs = expectJust "chooseExternalIds" $ lookupVarEnv bind_env id
613 (new_ids, show_unfold)
614 | omit_prags = ([], False)
615 | otherwise = addExternal id rhs
616 unfold_env' = extendVarEnv unfold_env id (name',show_unfold)
617 referrer' | isExportedId id = id
618 | otherwise = referrer
620 search (zip new_ids (repeat referrer') ++ rest) unfold_env' occ_env'
622 tidy_internal :: [Id] -> UnfoldEnv -> TidyOccEnv
623 -> IO (UnfoldEnv, TidyOccEnv)
624 tidy_internal [] unfold_env occ_env = return (unfold_env,occ_env)
625 tidy_internal (id:ids) unfold_env occ_env = do
626 (occ_env', name') <- tidyTopName mod nc_var Nothing occ_env id
627 let unfold_env' = extendVarEnv unfold_env id (name',False)
628 tidy_internal ids unfold_env' occ_env'
630 addExternal :: Id -> CoreExpr -> ([Id],Bool)
631 addExternal id rhs = (new_needed_ids, show_unfold)
633 new_needed_ids = unfold_ids ++
634 filter (not . (`elemVarSet` unfold_set))
635 (varSetElems worker_ids ++
636 varSetElems spec_ids) -- XXX non-det ordering
639 dont_inline = isNeverActive (inlinePragmaActivation (inlinePragInfo idinfo))
640 loop_breaker = isNonRuleLoopBreaker (occInfo idinfo)
641 bottoming_fn = isBottomingSig (newStrictnessInfo idinfo `orElse` topSig)
642 spec_ids = specInfoFreeVars (specInfo idinfo)
643 worker_info = workerInfo idinfo
645 -- Stuff to do with the Id's unfolding
646 -- The simplifier has put an up-to-date unfolding
647 -- in the IdInfo, but the RHS will do just as well
648 unfolding = unfoldingInfo idinfo
649 rhs_is_small = not (neverUnfold unfolding)
651 -- We leave the unfolding there even if there is a worker
652 -- In GHCI the unfolding is used by importers
653 -- When writing an interface file, we omit the unfolding
654 -- if there is a worker
655 show_unfold = not bottoming_fn && -- Not necessary
658 rhs_is_small -- Small enough
660 (unfold_set, unfold_ids)
661 | show_unfold = freeVarsInDepthFirstOrder rhs
662 | otherwise = (emptyVarSet, [])
664 worker_ids = case worker_info of
665 HasWorker work_id _ -> unitVarSet work_id
666 _otherwise -> emptyVarSet
669 -- We want a deterministic free-variable list. exprFreeVars gives us
670 -- a VarSet, which is in a non-deterministic order when converted to a
671 -- list. Hence, here we define a free-variable finder that returns
672 -- the free variables in the order that they are encountered.
674 -- Note [choosing external names]
676 freeVarsInDepthFirstOrder :: CoreExpr -> (VarSet, [Id])
677 freeVarsInDepthFirstOrder e =
679 DFFV m -> case m emptyVarSet [] of
680 (set,ids,_) -> (set,ids)
682 newtype DFFV a = DFFV (VarSet -> [Var] -> (VarSet,[Var],a))
684 instance Monad DFFV where
685 return a = DFFV $ \set ids -> (set, ids, a)
686 (DFFV m) >>= k = DFFV $ \set ids ->
688 (set',ids',a) -> case k a of
689 DFFV f -> f set' ids'
691 insert :: Var -> DFFV ()
692 insert v = DFFV $ \ set ids -> case () of
693 _ | v `elemVarSet` set -> (set,ids,())
694 | otherwise -> (extendVarSet set v, v:ids, ())
696 dffvExpr :: CoreExpr -> DFFV ()
697 dffvExpr e = go emptyVarSet e
699 go scope e = case e of
700 Var v | isLocalId v && not (v `elemVarSet` scope) -> insert v
701 App e1 e2 -> do go scope e1; go scope e2
702 Lam v e -> go (extendVarSet scope v) e
703 Note _ e -> go scope e
704 Cast e _ -> go scope e
705 Let (NonRec x r) e -> do go scope r; go (extendVarSet scope x) e
706 Let (Rec prs) e -> do let scope' = extendVarSetList scope (map fst prs)
707 mapM_ (go scope') (map snd prs)
709 Case e b _ as -> do go scope e
710 mapM_ (go_alt (extendVarSet scope b)) as
713 go_alt scope (_,xs,r) = go (extendVarSetList scope xs) r
717 --------------------------------------------------------------------
719 -- This is where we set names to local/global based on whether they really are
720 -- externally visible (see comment at the top of this module). If the name
721 -- was previously local, we have to give it a unique occurrence name if
722 -- we intend to externalise it.
725 tidyTopName :: Module -> IORef NameCache -> Maybe Id -> TidyOccEnv
726 -> Id -> IO (TidyOccEnv, Name)
727 tidyTopName mod nc_var maybe_ref occ_env id
728 | global && internal = return (occ_env, localiseName name)
730 | global && external = return (occ_env, name)
731 -- Global names are assumed to have been allocated by the renamer,
732 -- so they already have the "right" unique
733 -- And it's a system-wide unique too
735 -- Now we get to the real reason that all this is in the IO Monad:
736 -- we have to update the name cache in a nice atomic fashion
738 | local && internal = do { nc <- readIORef nc_var
739 ; let (nc', new_local_name) = mk_new_local nc
740 ; writeIORef nc_var nc'
741 ; return (occ_env', new_local_name) }
742 -- Even local, internal names must get a unique occurrence, because
743 -- if we do -split-objs we externalise the name later, in the code generator
745 -- Similarly, we must make sure it has a system-wide Unique, because
746 -- the byte-code generator builds a system-wide Name->BCO symbol table
748 | local && external = do { nc <- readIORef nc_var
749 ; let (nc', new_external_name) = mk_new_external nc
750 ; writeIORef nc_var nc'
751 ; return (occ_env', new_external_name) }
753 | otherwise = panic "tidyTopName"
756 external = isJust maybe_ref
757 global = isExternalName name
759 internal = not external
760 loc = nameSrcSpan name
762 old_occ = nameOccName name
764 | Just ref <- maybe_ref, ref /= id =
765 mkOccName (occNameSpace old_occ) $
766 occNameString (getOccName ref) ++ '_' : occNameString old_occ
767 | otherwise = old_occ
769 (occ_env', occ') = tidyOccName occ_env new_occ
771 mk_new_local nc = (nc { nsUniqs = us2 }, mkInternalName uniq occ' loc)
773 (us1, us2) = splitUniqSupply (nsUniqs nc)
774 uniq = uniqFromSupply us1
776 mk_new_external nc = allocateGlobalBinder nc mod occ' loc
777 -- If we want to externalise a currently-local name, check
778 -- whether we have already assigned a unique for it.
779 -- If so, use it; if not, extend the table.
780 -- All this is done by allcoateGlobalBinder.
781 -- This is needed when *re*-compiling a module in GHCi; we must
782 -- use the same name for externally-visible things as we did before.
786 findExternalRules :: [CoreBind]
787 -> [CoreRule] -- Non-local rules (i.e. ones for imported fns)
788 -> UnfoldEnv -- Ids that are exported, so we need their rules
790 -- The complete rules are gotten by combining
791 -- a) the non-local rules
792 -- b) rules embedded in the top-level Ids
793 findExternalRules binds non_local_rules unfold_env
794 = filter (not . internal_rule) (non_local_rules ++ local_rules)
797 | id <- bindersOfBinds binds,
799 rule <- idCoreRules id
803 = any (not . external_id) (varSetElems (ruleLhsFreeIds rule))
804 -- Don't export a rule whose LHS mentions a locally-defined
805 -- Id that is completely internal (i.e. not visible to an
809 | Just (name,_) <- lookupVarEnv unfold_env id = isExternalName name
815 %************************************************************************
817 \subsection{Step 2: top-level tidying}
819 %************************************************************************
823 -- TopTidyEnv: when tidying we need to know
824 -- * nc_var: The NameCache, containing a unique supply and any pre-ordained Names.
825 -- These may have arisen because the
826 -- renamer read in an interface file mentioning M.$wf, say,
827 -- and assigned it unique r77. If, on this compilation, we've
828 -- invented an Id whose name is $wf (but with a different unique)
829 -- we want to rename it to have unique r77, so that we can do easy
830 -- comparisons with stuff from the interface file
832 -- * occ_env: The TidyOccEnv, which tells us which local occurrences
835 -- * subst_env: A Var->Var mapping that substitutes the new Var for the old
837 tidyTopBinds :: HscEnv
841 -> (TidyEnv, [CoreBind])
843 tidyTopBinds hsc_env unfold_env init_occ_env binds
844 = tidy init_env binds
846 init_env = (init_occ_env, emptyVarEnv)
848 this_pkg = thisPackage (hsc_dflags hsc_env)
850 tidy env [] = (env, [])
851 tidy env (b:bs) = let (env1, b') = tidyTopBind this_pkg unfold_env env b
852 (env2, bs') = tidy env1 bs
856 ------------------------
857 tidyTopBind :: PackageId
861 -> (TidyEnv, CoreBind)
863 tidyTopBind this_pkg unfold_env (occ_env1,subst1) (NonRec bndr rhs)
864 = (tidy_env2, NonRec bndr' rhs')
866 Just (name',show_unfold) = lookupVarEnv unfold_env bndr
867 caf_info = hasCafRefs this_pkg subst1 (idArity bndr) rhs
868 (bndr', rhs') = tidyTopPair show_unfold tidy_env2 caf_info name' (bndr, rhs)
869 subst2 = extendVarEnv subst1 bndr bndr'
870 tidy_env2 = (occ_env1, subst2)
872 tidyTopBind this_pkg unfold_env (occ_env1,subst1) (Rec prs)
873 = (tidy_env2, Rec prs')
875 prs' = [ tidyTopPair show_unfold tidy_env2 caf_info name' (id,rhs)
877 let (name',show_unfold) =
878 expectJust "tidyTopBind" $ lookupVarEnv unfold_env id
881 subst2 = extendVarEnvList subst1 (bndrs `zip` map fst prs')
882 tidy_env2 = (occ_env1, subst2)
886 -- the CafInfo for a recursive group says whether *any* rhs in
887 -- the group may refer indirectly to a CAF (because then, they all do).
889 | or [ mayHaveCafRefs (hasCafRefs this_pkg subst1 (idArity bndr) rhs)
890 | (bndr,rhs) <- prs ] = MayHaveCafRefs
891 | otherwise = NoCafRefs
893 -----------------------------------------------------------
894 tidyTopPair :: Bool -- show unfolding
895 -> TidyEnv -- The TidyEnv is used to tidy the IdInfo
896 -- It is knot-tied: don't look at it!
899 -> (Id, CoreExpr) -- Binder and RHS before tidying
901 -- This function is the heart of Step 2
902 -- The rec_tidy_env is the one to use for the IdInfo
903 -- It's necessary because when we are dealing with a recursive
904 -- group, a variable late in the group might be mentioned
905 -- in the IdInfo of one early in the group
907 tidyTopPair show_unfold rhs_tidy_env caf_info name' (bndr, rhs)
910 bndr' = mkGlobalId details name' ty' idinfo'
911 details = idDetails bndr -- Preserve the IdDetails
912 ty' = tidyTopType (idType bndr)
913 rhs' = tidyExpr rhs_tidy_env rhs
915 idinfo' = tidyTopIdInfo (isExternalName name')
916 idinfo unfold_info worker_info
919 unfold_info | show_unfold = mkTopUnfolding rhs'
920 | otherwise = noUnfolding
921 worker_info = tidyWorker rhs_tidy_env show_unfold (workerInfo idinfo)
923 -- Usually the Id will have an accurate arity on it, because
924 -- the simplifier has just run, but not always.
925 -- One case I found was when the last thing the simplifier
926 -- did was to let-bind a non-atomic argument and then float
927 -- it to the top level. So it seems more robust just to
929 arity = exprArity rhs
932 -- tidyTopIdInfo creates the final IdInfo for top-level
933 -- binders. There are two delicate pieces:
935 -- * Arity. After CoreTidy, this arity must not change any more.
936 -- Indeed, CorePrep must eta expand where necessary to make
937 -- the manifest arity equal to the claimed arity.
939 -- * CAF info. This must also remain valid through to code generation.
940 -- We add the info here so that it propagates to all
941 -- occurrences of the binders in RHSs, and hence to occurrences in
942 -- unfoldings, which are inside Ids imported by GHCi. Ditto RULES.
943 -- CoreToStg makes use of this when constructing SRTs.
944 tidyTopIdInfo :: Bool -> IdInfo -> Unfolding
945 -> WorkerInfo -> ArityInfo -> CafInfo
947 tidyTopIdInfo is_external idinfo unfold_info worker_info arity caf_info
948 | not is_external -- For internal Ids (not externally visible)
949 = vanillaIdInfo -- we only need enough info for code generation
950 -- Arity and strictness info are enough;
951 -- c.f. CoreTidy.tidyLetBndr
952 `setCafInfo` caf_info
954 `setAllStrictnessInfo` newStrictnessInfo idinfo
956 | otherwise -- Externally-visible Ids get the whole lot
958 `setCafInfo` caf_info
960 `setAllStrictnessInfo` newStrictnessInfo idinfo
961 `setInlinePragInfo` inlinePragInfo idinfo
962 `setUnfoldingInfo` unfold_info
963 `setWorkerInfo` worker_info
964 -- NB: we throw away the Rules
965 -- They have already been extracted by findExternalRules
969 ------------ Worker --------------
970 tidyWorker :: TidyEnv -> Bool -> WorkerInfo -> WorkerInfo
971 tidyWorker _tidy_env _show_unfold NoWorker
973 tidyWorker tidy_env show_unfold (HasWorker work_id wrap_arity)
974 | show_unfold = HasWorker (tidyVarOcc tidy_env work_id) wrap_arity
975 | otherwise = NoWorker
976 -- NB: do *not* expose the worker if show_unfold is off,
977 -- because that means this thing is a loop breaker or
978 -- marked NOINLINE or something like that
979 -- This is important: if you expose the worker for a loop-breaker
980 -- then you can make the simplifier go into an infinite loop, because
981 -- in effect the unfolding is exposed. See Trac #1709
983 -- You might think that if show_unfold is False, then the thing should
984 -- not be w/w'd in the first place. But a legitimate reason is this:
985 -- the function returns bottom
986 -- In this case, show_unfold will be false (we don't expose unfoldings
987 -- for bottoming functions), but we might still have a worker/wrapper
988 -- split (see Note [Worker-wrapper for bottoming functions] in WorkWrap.lhs
991 %************************************************************************
993 \subsection{Figuring out CafInfo for an expression}
995 %************************************************************************
997 hasCafRefs decides whether a top-level closure can point into the dynamic heap.
998 We mark such things as `MayHaveCafRefs' because this information is
999 used to decide whether a particular closure needs to be referenced
1002 There are two reasons for setting MayHaveCafRefs:
1003 a) The RHS is a CAF: a top-level updatable thunk.
1004 b) The RHS refers to something that MayHaveCafRefs
1006 Possible improvement: In an effort to keep the number of CAFs (and
1007 hence the size of the SRTs) down, we could also look at the expression and
1008 decide whether it requires a small bounded amount of heap, so we can ignore
1009 it as a CAF. In these cases however, we would need to use an additional
1010 CAF list to keep track of non-collectable CAFs.
1013 hasCafRefs :: PackageId -> VarEnv Var -> Arity -> CoreExpr -> CafInfo
1014 hasCafRefs this_pkg p arity expr
1015 | is_caf || mentions_cafs
1017 | otherwise = NoCafRefs
1019 mentions_cafs = isFastTrue (cafRefs p expr)
1020 is_caf = not (arity > 0 || rhsIsStatic this_pkg expr)
1022 -- NB. we pass in the arity of the expression, which is expected
1023 -- to be calculated by exprArity. This is because exprArity
1024 -- knows how much eta expansion is going to be done by
1025 -- CorePrep later on, and we don't want to duplicate that
1026 -- knowledge in rhsIsStatic below.
1028 cafRefs :: VarEnv Id -> Expr a -> FastBool
1030 -- imported Ids first:
1031 | not (isLocalId id) = fastBool (mayHaveCafRefs (idCafInfo id))
1032 -- now Ids local to this module:
1034 case lookupVarEnv p id of
1035 Just id' -> fastBool (mayHaveCafRefs (idCafInfo id'))
1036 Nothing -> fastBool False
1038 cafRefs _ (Lit _) = fastBool False
1039 cafRefs p (App f a) = fastOr (cafRefs p f) (cafRefs p) a
1040 cafRefs p (Lam _ e) = cafRefs p e
1041 cafRefs p (Let b e) = fastOr (cafRefss p (rhssOfBind b)) (cafRefs p) e
1042 cafRefs p (Case e _bndr _ alts) = fastOr (cafRefs p e) (cafRefss p) (rhssOfAlts alts)
1043 cafRefs p (Note _n e) = cafRefs p e
1044 cafRefs p (Cast e _co) = cafRefs p e
1045 cafRefs _ (Type _) = fastBool False
1047 cafRefss :: VarEnv Id -> [Expr a] -> FastBool
1048 cafRefss _ [] = fastBool False
1049 cafRefss p (e:es) = fastOr (cafRefs p e) (cafRefss p) es
1051 fastOr :: FastBool -> (a -> FastBool) -> a -> FastBool
1052 -- hack for lazy-or over FastBool.
1053 fastOr a f x = fastBool (isFastTrue a || isFastTrue (f x))