2 % (c) The GRASP/AQUA Project, Glasgow University, 1992-1998
4 \section{Tidying up Core}
7 module TidyPgm( mkBootModDetailsDs, mkBootModDetailsTc, tidyProgram ) where
9 #include "HsVersions.h"
44 import FastBool hiding ( fastOr )
46 import Data.List ( partition )
47 import Data.Maybe ( isJust )
48 import Data.IORef ( IORef, readIORef, writeIORef )
52 Constructing the TypeEnv, Instances, Rules from which the ModIface is
53 constructed, and which goes on to subsequent modules in --make mode.
55 Most of the interface file is obtained simply by serialising the
56 TypeEnv. One important consequence is that if the *interface file*
57 has pragma info if and only if the final TypeEnv does. This is not so
58 important for *this* module, but it's essential for ghc --make:
59 subsequent compilations must not see (e.g.) the arity if the interface
60 file does not contain arity If they do, they'll exploit the arity;
61 then the arity might change, but the iface file doesn't change =>
62 recompilation does not happen => disaster.
64 For data types, the final TypeEnv will have a TyThing for the TyCon,
65 plus one for each DataCon; the interface file will contain just one
66 data type declaration, but it is de-serialised back into a collection
69 %************************************************************************
73 %************************************************************************
76 Plan A: mkBootModDetails: omit pragmas, make interfaces small
77 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
80 * Drop all WiredIn things from the TypeEnv
81 (we never want them in interface files)
83 * Retain all TyCons and Classes in the TypeEnv, to avoid
84 having to find which ones are mentioned in the
87 * Trim off the constructors of non-exported TyCons, both
88 from the TyCon and from the TypeEnv
90 * Drop non-exported Ids from the TypeEnv
92 * Tidy the types of the DFunIds of Instances,
93 make them into GlobalIds, (they already have External Names)
94 and add them to the TypeEnv
96 * Tidy the types of the (exported) Ids in the TypeEnv,
97 make them into GlobalIds (they already have External Names)
99 * Drop rules altogether
101 * Tidy the bindings, to ensure that the Caf and Arity
102 information is correct for each top-level binder; the
103 code generator needs it. And to ensure that local names have
104 distinct OccNames in case of object-file splitting
107 -- This is Plan A: make a small type env when typechecking only,
108 -- or when compiling a hs-boot file, or simply when not using -O
110 -- We don't look at the bindings at all -- there aren't any
113 mkBootModDetailsTc :: HscEnv -> TcGblEnv -> IO ModDetails
114 mkBootModDetailsTc hsc_env
115 TcGblEnv{ tcg_exports = exports,
116 tcg_type_env = type_env,
118 tcg_fam_insts = fam_insts
120 = mkBootModDetails hsc_env exports type_env insts fam_insts
122 mkBootModDetailsDs :: HscEnv -> ModGuts -> IO ModDetails
123 mkBootModDetailsDs hsc_env
124 ModGuts{ mg_exports = exports,
127 mg_fam_insts = fam_insts
129 = mkBootModDetails hsc_env exports type_env insts fam_insts
131 mkBootModDetails :: HscEnv -> [AvailInfo] -> NameEnv TyThing
132 -> [Instance] -> [FamInstEnv.FamInst] -> IO ModDetails
133 mkBootModDetails hsc_env exports type_env insts fam_insts
134 = do { let dflags = hsc_dflags hsc_env
135 ; showPass dflags "Tidy [hoot] type env"
137 ; let { insts' = tidyInstances tidyExternalId insts
138 ; dfun_ids = map instanceDFunId insts'
139 ; type_env1 = tidyBootTypeEnv (availsToNameSet exports) type_env
140 ; type_env' = extendTypeEnvWithIds type_env1 dfun_ids
142 ; return (ModDetails { md_types = type_env'
144 , md_fam_insts = fam_insts
147 , md_exports = exports
148 , md_vect_info = noVectInfo
153 tidyBootTypeEnv :: NameSet -> TypeEnv -> TypeEnv
154 tidyBootTypeEnv exports type_env
155 = tidyTypeEnv True False exports type_env final_ids
157 -- Find the LocalIds in the type env that are exported
158 -- Make them into GlobalIds, and tidy their types
160 -- It's very important to remove the non-exported ones
161 -- because we don't tidy the OccNames, and if we don't remove
162 -- the non-exported ones we'll get many things with the
163 -- same name in the interface file, giving chaos.
164 final_ids = [ tidyExternalId id
165 | id <- typeEnvIds type_env
169 -- default methods have their export flag set, but everything
170 -- else doesn't (yet), because this is pre-desugaring, so we
172 keep_it id = isExportedId id || idName id `elemNameSet` exports
175 tidyExternalId :: Id -> Id
176 -- Takes an LocalId with an External Name,
177 -- makes it into a GlobalId with VanillaIdInfo, and tidies its type
178 -- (NB: vanillaIdInfo makes a conservative assumption about Caf-hood.)
180 = ASSERT2( isLocalId id && isExternalName (idName id), ppr id )
181 mkVanillaGlobal (idName id) (tidyTopType (idType id))
185 %************************************************************************
187 Plan B: tidy bindings, make TypeEnv full of IdInfo
189 %************************************************************************
191 Plan B: include pragmas, make interfaces
192 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
193 * Figure out which Ids are externally visible
195 * Tidy the bindings, externalising appropriate Ids
197 * Drop all Ids from the TypeEnv, and add all the External Ids from
198 the bindings. (This adds their IdInfo to the TypeEnv; and adds
199 floated-out Ids that weren't even in the TypeEnv before.)
201 Step 1: Figure out external Ids
202 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
203 First we figure out which Ids are "external" Ids. An
204 "external" Id is one that is visible from outside the compilation
206 a) the user exported ones
207 b) ones mentioned in the unfoldings, workers,
208 or rules of externally-visible ones
209 This exercise takes a sweep of the bindings bottom to top. Actually,
210 in Step 2 we're also going to need to know which Ids should be
211 exported with their unfoldings, so we produce not an IdSet but an
212 ExtIdEnv = IdEnv Bool
215 Step 2: Tidy the program
216 ~~~~~~~~~~~~~~~~~~~~~~~~
217 Next we traverse the bindings top to bottom. For each *top-level*
220 1. Make it into a GlobalId; its IdDetails becomes VanillaGlobal,
221 reflecting the fact that from now on we regard it as a global,
224 2. Give it a system-wide Unique.
225 [Even non-exported things need system-wide Uniques because the
226 byte-code generator builds a single Name->BCO symbol table.]
228 We use the NameCache kept in the HscEnv as the
229 source of such system-wide uniques.
231 For external Ids, use the original-name cache in the NameCache
232 to ensure that the unique assigned is the same as the Id had
233 in any previous compilation run.
235 3. If it's an external Id, make it have a External Name, otherwise
236 make it have an Internal Name.
237 This is used by the code generator to decide whether
238 to make the label externally visible
240 4. Give external Ids a "tidy" OccName. This means
241 we can print them in interface files without confusing
242 "x" (unique 5) with "x" (unique 10).
244 5. Give it its UTTERLY FINAL IdInfo; in ptic,
245 * its unfolding, if it should have one
247 * its arity, computed from the number of visible lambdas
249 * its CAF info, computed from what is free in its RHS
252 Finally, substitute these new top-level binders consistently
253 throughout, including in unfoldings. We also tidy binders in
254 RHSs, so that they print nicely in interfaces.
257 tidyProgram :: HscEnv -> ModGuts -> IO (CgGuts, ModDetails)
258 tidyProgram hsc_env (ModGuts { mg_module = mod, mg_exports = exports,
260 mg_insts = insts, mg_fam_insts = fam_insts,
262 mg_rules = imp_rules,
263 mg_vect_info = vect_info,
264 mg_dir_imps = dir_imps,
267 mg_foreign = foreign_stubs,
268 mg_hpc_info = hpc_info,
269 mg_modBreaks = modBreaks })
271 = do { let dflags = hsc_dflags hsc_env
272 ; showPass dflags "Tidy Core"
274 ; let { omit_prags = dopt Opt_OmitInterfacePragmas dflags
275 ; th = dopt Opt_TemplateHaskell dflags
276 ; ext_ids = findExternalIds omit_prags binds
279 | otherwise = findExternalRules binds imp_rules ext_ids
280 -- findExternalRules filters imp_rules to avoid binders that
281 -- aren't externally visible; but the externally-visible binders
282 -- are computed (by findExternalIds) assuming that all orphan
283 -- rules are exported (they get their Exported flag set in the desugarer)
284 -- So in fact we may export more than we need.
285 -- (It's a sort of mutual recursion.)
288 ; (tidy_env, tidy_binds) <- tidyTopBinds hsc_env mod type_env ext_ids
291 ; let { export_set = availsToNameSet exports
292 ; final_ids = [ id | id <- bindersOfBinds tidy_binds,
293 isExternalName (idName id)]
294 ; tidy_type_env = tidyTypeEnv omit_prags th export_set
296 ; tidy_insts = tidyInstances (lookup_dfun tidy_type_env) insts
297 -- A DFunId will have a binding in tidy_binds, and so
298 -- will now be in final_env, replete with IdInfo
299 -- Its name will be unchanged since it was born, but
300 -- we want Global, IdInfo-rich (or not) DFunId in the
303 ; tidy_rules = tidyRules tidy_env ext_rules
304 -- You might worry that the tidy_env contains IdInfo-rich stuff
305 -- and indeed it does, but if omit_prags is on, ext_rules is
308 -- See Note [Injecting implicit bindings]
309 ; implicit_binds = getImplicitBinds type_env
310 ; all_tidy_binds = implicit_binds ++ tidy_binds
312 ; alg_tycons = filter isAlgTyCon (typeEnvTyCons type_env)
315 ; endPass dflags "Tidy Core" Opt_D_dump_simpl all_tidy_binds
316 ; dumpIfSet_core dflags Opt_D_dump_simpl
318 (pprRules tidy_rules)
320 ; let dir_imp_mods = moduleEnvKeys dir_imps
322 ; return (CgGuts { cg_module = mod,
323 cg_tycons = alg_tycons,
324 cg_binds = all_tidy_binds,
325 cg_dir_imps = dir_imp_mods,
326 cg_foreign = foreign_stubs,
327 cg_dep_pkgs = dep_pkgs deps,
328 cg_hpc_info = hpc_info,
329 cg_modBreaks = modBreaks },
331 ModDetails { md_types = tidy_type_env,
332 md_rules = tidy_rules,
333 md_insts = tidy_insts,
334 md_fam_insts = fam_insts,
335 md_exports = exports,
336 md_anns = anns, -- are already tidy
337 md_vect_info = vect_info --
341 lookup_dfun :: TypeEnv -> Var -> Id
342 lookup_dfun type_env dfun_id
343 = case lookupTypeEnv type_env (idName dfun_id) of
344 Just (AnId dfun_id') -> dfun_id'
345 _other -> pprPanic "lookup_dfun" (ppr dfun_id)
347 --------------------------
348 tidyTypeEnv :: Bool -- Compiling without -O, so omit prags
349 -> Bool -- Template Haskell is on
350 -> NameSet -> TypeEnv -> [Id] -> TypeEnv
352 -- The competed type environment is gotten from
353 -- Dropping any wired-in things, and then
354 -- a) keeping the types and classes
355 -- b) removing all Ids,
356 -- c) adding Ids with correct IdInfo, including unfoldings,
357 -- gotten from the bindings
358 -- From (c) we keep only those Ids with External names;
359 -- the CoreTidy pass makes sure these are all and only
360 -- the externally-accessible ones
361 -- This truncates the type environment to include only the
362 -- exported Ids and things needed from them, which saves space
364 tidyTypeEnv omit_prags th exports type_env final_ids
365 = let type_env1 = filterNameEnv keep_it type_env
366 type_env2 = extendTypeEnvWithIds type_env1 final_ids
367 type_env3 | omit_prags = mapNameEnv (trimThing th exports) type_env2
368 | otherwise = type_env2
372 -- We keep GlobalIds, because they won't appear
373 -- in the bindings from which final_ids are derived!
374 -- (The bindings bind LocalIds.)
375 keep_it thing | isWiredInThing thing = False
376 keep_it (AnId id) = isGlobalId id -- Keep GlobalIds (e.g. class ops)
377 keep_it _other = True -- Keep all TyCons, DataCons, and Classes
379 --------------------------
380 isWiredInThing :: TyThing -> Bool
381 isWiredInThing thing = isWiredInName (getName thing)
383 --------------------------
384 trimThing :: Bool -> NameSet -> TyThing -> TyThing
385 -- Trim off inessentials, for boot files and no -O
386 trimThing th exports (ATyCon tc)
387 | not th && not (mustExposeTyCon exports tc)
388 = ATyCon (makeTyConAbstract tc) -- Note [Trimming and Template Haskell]
390 trimThing _th _exports (AnId id)
391 | not (isImplicitId id)
392 = AnId (id `setIdInfo` vanillaIdInfo)
394 trimThing _th _exports other_thing
398 {- Note [Trimming and Template Haskell]
399 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
400 Consider (Trac #2386) this
401 module M(T, makeOne) where
403 makeOne = [| Yay "Yep" |]
404 Notice that T is exported abstractly, but makeOne effectively exports it too!
405 A module that splices in $(makeOne) will then look for a declartion of Yay,
406 so it'd better be there. Hence, brutally but simply, we switch off type
407 constructor trimming if TH is enabled in this module. -}
410 mustExposeTyCon :: NameSet -- Exports
411 -> TyCon -- The tycon
412 -> Bool -- Can its rep be hidden?
413 -- We are compiling without -O, and thus trying to write as little as
414 -- possible into the interface file. But we must expose the details of
415 -- any data types whose constructors or fields are exported
416 mustExposeTyCon exports tc
417 | not (isAlgTyCon tc) -- Synonyms
419 | isEnumerationTyCon tc -- For an enumeration, exposing the constructors
420 = True -- won't lead to the need for further exposure
421 -- (This includes data types with no constructors.)
422 | isOpenTyCon tc -- Open type family
425 | otherwise -- Newtype, datatype
426 = any exported_con (tyConDataCons tc)
427 -- Expose rep if any datacon or field is exported
429 || (isNewTyCon tc && isFFITy (snd (newTyConRhs tc)))
430 -- Expose the rep for newtypes if the rep is an FFI type.
431 -- For a very annoying reason. 'Foreign import' is meant to
432 -- be able to look through newtypes transparently, but it
433 -- can only do that if it can "see" the newtype representation
435 exported_con con = any (`elemNameSet` exports)
436 (dataConName con : dataConFieldLabels con)
438 tidyInstances :: (DFunId -> DFunId) -> [Instance] -> [Instance]
439 tidyInstances tidy_dfun ispecs
442 tidy ispec = setInstanceDFunId ispec $
443 tidy_dfun (instanceDFunId ispec)
447 %************************************************************************
451 %************************************************************************
453 Note [Injecting implicit bindings]
454 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
455 We inject the implict bindings right at the end, in CoreTidy.
456 Some of these bindings, notably record selectors, are not
457 constructed in an optimised form. E.g. record selector for
458 data T = MkT { x :: {-# UNPACK #-} !Int }
459 Then the unfolding looks like
460 x = \t. case t of MkT x1 -> let x = I# x1 in x
461 This generates bad code unless it's first simplified a bit. That is
462 why CoreUnfold.mkImplicitUnfolding uses simleExprOpt to do a bit of
463 optimisation first. (Only matters when the selector is used curried;
464 eg map x ys.) See Trac #2070.
466 At one time I tried injecting the implicit bindings *early*, at the
467 beginning of SimplCore. But that gave rise to real difficulty,
468 becuase GlobalIds are supposed to have *fixed* IdInfo, but the
469 simplifier and other core-to-core passes mess with IdInfo all the
470 time. The straw that broke the camels back was when a class selector
471 got the wrong arity -- ie the simplifier gave it arity 2, whereas
472 importing modules were expecting it to have arity 1 (Trac #2844).
473 It's much safer just to inject them right at the end, after tidying.
477 getImplicitBinds :: TypeEnv -> [CoreBind]
478 getImplicitBinds type_env
479 = map get_defn (concatMap implicit_con_ids (typeEnvTyCons type_env)
480 ++ concatMap other_implicit_ids (typeEnvElts type_env))
481 -- Put the constructor wrappers first, because
482 -- other implicit bindings (notably the fromT functions arising
483 -- from generics) use the constructor wrappers. At least that's
484 -- what External Core likes
486 implicit_con_ids tc = mapCatMaybes dataConWrapId_maybe (tyConDataCons tc)
488 other_implicit_ids (ATyCon tc) = filter (not . isNaughtyRecordSelector) (tyConSelIds tc)
489 -- The "naughty" ones are not real functions at all
490 -- They are there just so we can get decent error messages
491 -- See Note [Naughty record selectors] in MkId.lhs
492 other_implicit_ids (AClass cl) = classSelIds cl
493 other_implicit_ids _other = []
495 get_defn :: Id -> CoreBind
496 get_defn id = NonRec id (unfoldingTemplate (idUnfolding id))
500 %************************************************************************
502 \subsection{Step 1: finding externals}
504 %************************************************************************
507 type ExtIdEnv = IdEnv Bool
508 -- In domain => Id is external
509 -- Range = True <=> show unfolding,
510 -- Always True for InlineRule
512 findExternalIds :: Bool -> [CoreBind] -> ExtIdEnv
513 -- Step 1 from the notes above
514 findExternalIds omit_prags binds
516 = mkVarEnv [ (id,False) | id <- bindersOfBinds binds, isExportedId id ]
519 = foldr find emptyVarEnv binds
521 find (NonRec id rhs) needed
522 | need_id needed id = addExternal (id,rhs) needed
524 find (Rec prs) needed = find_prs prs needed
526 -- For a recursive group we have to look for a fixed point
528 | null needed_prs = needed
529 | otherwise = find_prs other_prs new_needed
531 (needed_prs, other_prs) = partition (need_pr needed) prs
532 new_needed = foldr addExternal needed needed_prs
534 -- The 'needed' set contains the Ids that are needed by earlier
535 -- interface file emissions. If the Id isn't in this set, and isn't
536 -- exported, there's no need to emit anything
537 need_id needed_set id = id `elemVarEnv` needed_set || isExportedId id
538 need_pr needed_set (id,_) = need_id needed_set id
540 addExternal :: (Id,CoreExpr) -> IdEnv Bool -> IdEnv Bool
541 -- The Id is needed; extend the needed set
542 -- with it and its dependents (free vars etc)
543 addExternal (id,rhs) needed
544 = extendVarEnv (foldVarSet add_occ needed new_needed_ids)
547 add_occ id needed | id `elemVarEnv` needed = needed
548 | otherwise = extendVarEnv needed id False
549 -- "False" because we don't know we need the Id's unfolding
550 -- Don't override existing bindings; we might have already set it to True
552 new_needed_ids = (mb_unfold_ids `orElse` emptyVarSet) `unionVarSet`
556 dont_inline = isNeverActive (inlinePragInfo idinfo)
557 loop_breaker = isNonRuleLoopBreaker (occInfo idinfo)
558 bottoming_fn = isBottomingSig (newStrictnessInfo idinfo `orElse` topSig)
559 spec_ids = specInfoFreeVars (specInfo idinfo)
561 -- Stuff to do with the Id's unfolding
562 -- We leave the unfolding there even if there is a worker
563 -- In GHCI the unfolding is used by importers
564 show_unfold = isJust mb_unfold_ids
566 mb_unfold_ids :: Maybe IdSet -- Nothing => don't unfold
567 mb_unfold_ids = case unfoldingInfo idinfo of
568 InlineRule { uf_worker = Just wkr_id } -> Just (unitVarSet wkr_id)
569 InlineRule { uf_tmpl = rhs } -> Just (exprFreeIds rhs)
570 CoreUnfolding { uf_guidance = guide }
571 | not bottoming_fn -- Not necessary
574 , not (neverUnfoldGuidance guide)
575 -> Just (exprFreeIds rhs) -- The simplifier has put an up-to-date unfolding
576 -- in the IdInfo, but the RHS will do just as well
583 findExternalRules :: [CoreBind]
584 -> [CoreRule] -- Non-local rules (i.e. ones for imported fns)
585 -> IdEnv a -- Ids that are exported, so we need their rules
587 -- The complete rules are gotten by combining
588 -- a) the non-local rules
589 -- b) rules embedded in the top-level Ids
590 findExternalRules binds non_local_rules ext_ids
591 = filter (not . internal_rule) (non_local_rules ++ local_rules)
594 | id <- bindersOfBinds binds,
595 id `elemVarEnv` ext_ids,
596 rule <- idCoreRules id
600 = any internal_id (varSetElems (ruleLhsFreeIds rule))
601 -- Don't export a rule whose LHS mentions a locally-defined
602 -- Id that is completely internal (i.e. not visible to an
605 internal_id id = not (id `elemVarEnv` ext_ids)
610 %************************************************************************
612 \subsection{Step 2: top-level tidying}
614 %************************************************************************
618 -- TopTidyEnv: when tidying we need to know
619 -- * nc_var: The NameCache, containing a unique supply and any pre-ordained Names.
620 -- These may have arisen because the
621 -- renamer read in an interface file mentioning M.$wf, say,
622 -- and assigned it unique r77. If, on this compilation, we've
623 -- invented an Id whose name is $wf (but with a different unique)
624 -- we want to rename it to have unique r77, so that we can do easy
625 -- comparisons with stuff from the interface file
627 -- * occ_env: The TidyOccEnv, which tells us which local occurrences
630 -- * subst_env: A Var->Var mapping that substitutes the new Var for the old
632 tidyTopBinds :: HscEnv
637 -> IO (TidyEnv, [CoreBind])
639 tidyTopBinds hsc_env mod type_env ext_ids binds
640 = tidy init_env binds
642 nc_var = hsc_NC hsc_env
644 -- We also make sure to avoid any exported binders. Consider
645 -- f{-u1-} = 1 -- Local decl
647 -- f{-u2-} = 2 -- Exported decl
649 -- The second exported decl must 'get' the name 'f', so we
650 -- have to put 'f' in the avoids list before we get to the first
651 -- decl. tidyTopId then does a no-op on exported binders.
652 init_env = (initTidyOccEnv avoids, emptyVarEnv)
653 avoids = [getOccName name | bndr <- typeEnvIds type_env,
654 let name = idName bndr,
656 -- In computing our "avoids" list, we must include
658 -- all things with global names (assigned once and for
659 -- all by the renamer)
660 -- since their names are "taken".
661 -- The type environment is a convenient source of such things.
663 this_pkg = thisPackage (hsc_dflags hsc_env)
665 tidy env [] = return (env, [])
666 tidy env (b:bs) = do { (env1, b') <- tidyTopBind this_pkg mod nc_var ext_ids env b
667 ; (env2, bs') <- tidy env1 bs
668 ; return (env2, b':bs') }
670 ------------------------
671 tidyTopBind :: PackageId
673 -> IORef NameCache -- For allocating new unique names
675 -> TidyEnv -> CoreBind
676 -> IO (TidyEnv, CoreBind)
678 tidyTopBind this_pkg mod nc_var ext_ids (occ_env1,subst1) (NonRec bndr rhs)
679 = do { (occ_env2, name') <- tidyTopName mod nc_var ext_ids occ_env1 bndr
680 ; let { (bndr', rhs') = tidyTopPair ext_ids tidy_env2 caf_info name' (bndr, rhs)
681 ; subst2 = extendVarEnv subst1 bndr bndr'
682 ; tidy_env2 = (occ_env2, subst2) }
683 ; return (tidy_env2, NonRec bndr' rhs') }
685 caf_info = hasCafRefs this_pkg subst1 (idArity bndr) rhs
687 tidyTopBind this_pkg mod nc_var ext_ids (occ_env1,subst1) (Rec prs)
688 = do { (occ_env2, names') <- tidyTopNames mod nc_var ext_ids occ_env1 bndrs
689 ; let { prs' = zipWith (tidyTopPair ext_ids tidy_env2 caf_info)
691 ; subst2 = extendVarEnvList subst1 (bndrs `zip` map fst prs')
692 ; tidy_env2 = (occ_env2, subst2) }
693 ; return (tidy_env2, Rec prs') }
697 -- the CafInfo for a recursive group says whether *any* rhs in
698 -- the group may refer indirectly to a CAF (because then, they all do).
700 | or [ mayHaveCafRefs (hasCafRefs this_pkg subst1 (idArity bndr) rhs)
701 | (bndr,rhs) <- prs ] = MayHaveCafRefs
702 | otherwise = NoCafRefs
704 --------------------------------------------------------------------
706 -- This is where we set names to local/global based on whether they really are
707 -- externally visible (see comment at the top of this module). If the name
708 -- was previously local, we have to give it a unique occurrence name if
709 -- we intend to externalise it.
710 tidyTopNames :: Module -> IORef NameCache -> VarEnv Bool -> TidyOccEnv
711 -> [Id] -> IO (TidyOccEnv, [Name])
712 tidyTopNames _mod _nc_var _ext_ids occ_env [] = return (occ_env, [])
713 tidyTopNames mod nc_var ext_ids occ_env (id:ids)
714 = do { (occ_env1, name) <- tidyTopName mod nc_var ext_ids occ_env id
715 ; (occ_env2, names) <- tidyTopNames mod nc_var ext_ids occ_env1 ids
716 ; return (occ_env2, name:names) }
718 tidyTopName :: Module -> IORef NameCache -> VarEnv Bool -> TidyOccEnv
719 -> Id -> IO (TidyOccEnv, Name)
720 tidyTopName mod nc_var ext_ids occ_env id
721 | global && internal = return (occ_env, localiseName name)
723 | global && external = return (occ_env, name)
724 -- Global names are assumed to have been allocated by the renamer,
725 -- so they already have the "right" unique
726 -- And it's a system-wide unique too
728 -- Now we get to the real reason that all this is in the IO Monad:
729 -- we have to update the name cache in a nice atomic fashion
731 | local && internal = do { nc <- readIORef nc_var
732 ; let (nc', new_local_name) = mk_new_local nc
733 ; writeIORef nc_var nc'
734 ; return (occ_env', new_local_name) }
735 -- Even local, internal names must get a unique occurrence, because
736 -- if we do -split-objs we externalise the name later, in the code generator
738 -- Similarly, we must make sure it has a system-wide Unique, because
739 -- the byte-code generator builds a system-wide Name->BCO symbol table
741 | local && external = do { nc <- readIORef nc_var
742 ; let (nc', new_external_name) = mk_new_external nc
743 ; writeIORef nc_var nc'
744 ; return (occ_env', new_external_name) }
746 | otherwise = panic "tidyTopName"
749 external = id `elemVarEnv` ext_ids
750 global = isExternalName name
752 internal = not external
753 loc = nameSrcSpan name
755 (occ_env', occ') = tidyOccName occ_env (nameOccName name)
757 mk_new_local nc = (nc { nsUniqs = us2 }, mkInternalName uniq occ' loc)
759 (us1, us2) = splitUniqSupply (nsUniqs nc)
760 uniq = uniqFromSupply us1
762 mk_new_external nc = allocateGlobalBinder nc mod occ' loc
763 -- If we want to externalise a currently-local name, check
764 -- whether we have already assigned a unique for it.
765 -- If so, use it; if not, extend the table.
766 -- All this is done by allcoateGlobalBinder.
767 -- This is needed when *re*-compiling a module in GHCi; we must
768 -- use the same name for externally-visible things as we did before.
771 -----------------------------------------------------------
772 tidyTopPair :: VarEnv Bool
773 -> TidyEnv -- The TidyEnv is used to tidy the IdInfo
774 -- It is knot-tied: don't look at it!
777 -> (Id, CoreExpr) -- Binder and RHS before tidying
779 -- This function is the heart of Step 2
780 -- The rec_tidy_env is the one to use for the IdInfo
781 -- It's necessary because when we are dealing with a recursive
782 -- group, a variable late in the group might be mentioned
783 -- in the IdInfo of one early in the group
785 tidyTopPair ext_ids rhs_tidy_env caf_info name' (bndr, rhs)
788 bndr' = mkGlobalId details name' ty' idinfo'
789 -- Preserve the GlobalIdDetails of existing global-ids
790 details = case globalIdDetails bndr of
791 NotGlobalId -> VanillaGlobal
792 old_details -> old_details
793 ty' = tidyTopType (idType bndr)
794 rhs' = tidyExpr rhs_tidy_env rhs
796 idinfo' = tidyTopIdInfo (isJust maybe_external)
800 -- Expose an unfolding if ext_ids tells us to
801 -- Remember that ext_ids maps an Id to a Bool:
802 -- True to show the unfolding, False to hide it
803 maybe_external = lookupVarEnv ext_ids bndr
804 show_unfold = maybe_external `orElse` False
805 unfold_info | show_unfold = tidyUnfolding rhs_tidy_env rhs' (unfoldingInfo idinfo)
806 | otherwise = noUnfolding
807 -- NB: do *not* expose the worker if show_unfold is off,
808 -- because that means this thing is a loop breaker or
809 -- marked NOINLINE or something like that
810 -- This is important: if you expose the worker for a loop-breaker
811 -- then you can make the simplifier go into an infinite loop, because
812 -- in effect the unfolding is exposed. See Trac #1709
814 -- You might think that if show_unfold is False, then the thing should
815 -- not be w/w'd in the first place. But a legitimate reason is this:
816 -- the function returns bottom
817 -- In this case, show_unfold will be false (we don't expose unfoldings
818 -- for bottoming functions), but we might still have a worker/wrapper
819 -- split (see Note [Worker-wrapper for bottoming functions] in WorkWrap.lhs
821 -- Usually the Id will have an accurate arity on it, because
822 -- the simplifier has just run, but not always.
823 -- One case I found was when the last thing the simplifier
824 -- did was to let-bind a non-atomic argument and then float
825 -- it to the top level. So it seems more robust just to
827 arity = exprArity rhs
830 -- tidyTopIdInfo creates the final IdInfo for top-level
831 -- binders. There are two delicate pieces:
833 -- * Arity. After CoreTidy, this arity must not change any more.
834 -- Indeed, CorePrep must eta expand where necessary to make
835 -- the manifest arity equal to the claimed arity.
837 -- * CAF info. This must also remain valid through to code generation.
838 -- We add the info here so that it propagates to all
839 -- occurrences of the binders in RHSs, and hence to occurrences in
840 -- unfoldings, which are inside Ids imported by GHCi. Ditto RULES.
841 -- CoreToStg makes use of this when constructing SRTs.
842 tidyTopIdInfo :: Bool -> IdInfo -> Unfolding
843 -> ArityInfo -> CafInfo
845 tidyTopIdInfo is_external idinfo unfold_info arity caf_info
846 | not is_external -- For internal Ids (not externally visible)
847 = vanillaIdInfo -- we only need enough info for code generation
848 -- Arity and strictness info are enough;
849 -- c.f. CoreTidy.tidyLetBndr
850 `setCafInfo` caf_info
852 `setAllStrictnessInfo` newStrictnessInfo idinfo
854 | otherwise -- Externally-visible Ids get the whole lot
856 `setCafInfo` caf_info
858 `setAllStrictnessInfo` newStrictnessInfo idinfo
859 `setInlinePragInfo` inlinePragInfo idinfo
860 `setUnfoldingInfo` unfold_info
861 -- NB: we throw away the Rules
862 -- They have already been extracted by findExternalRules
866 ------------ Unfolding --------------
867 tidyUnfolding :: TidyEnv -> CoreExpr -> Unfolding -> Unfolding
868 tidyUnfolding tidy_env _ unf@(InlineRule { uf_tmpl = rhs, uf_worker = mb_wkr })
869 = unf { uf_tmpl = tidyExpr tidy_env rhs,
870 uf_worker = fmap (tidyVarOcc tidy_env) mb_wkr }
871 tidyUnfolding _ tidy_rhs (CoreUnfolding {})
872 = mkTopUnfolding tidy_rhs
873 tidyUnfolding _ _ unf = unf
876 %************************************************************************
878 \subsection{Figuring out CafInfo for an expression}
880 %************************************************************************
882 hasCafRefs decides whether a top-level closure can point into the dynamic heap.
883 We mark such things as `MayHaveCafRefs' because this information is
884 used to decide whether a particular closure needs to be referenced
887 There are two reasons for setting MayHaveCafRefs:
888 a) The RHS is a CAF: a top-level updatable thunk.
889 b) The RHS refers to something that MayHaveCafRefs
891 Possible improvement: In an effort to keep the number of CAFs (and
892 hence the size of the SRTs) down, we could also look at the expression and
893 decide whether it requires a small bounded amount of heap, so we can ignore
894 it as a CAF. In these cases however, we would need to use an additional
895 CAF list to keep track of non-collectable CAFs.
898 hasCafRefs :: PackageId -> VarEnv Var -> Arity -> CoreExpr -> CafInfo
899 hasCafRefs this_pkg p arity expr
900 | is_caf || mentions_cafs
902 | otherwise = NoCafRefs
904 mentions_cafs = isFastTrue (cafRefs p expr)
905 is_caf = not (arity > 0 || rhsIsStatic this_pkg expr)
907 -- NB. we pass in the arity of the expression, which is expected
908 -- to be calculated by exprArity. This is because exprArity
909 -- knows how much eta expansion is going to be done by
910 -- CorePrep later on, and we don't want to duplicate that
911 -- knowledge in rhsIsStatic below.
913 cafRefs :: VarEnv Id -> Expr a -> FastBool
915 -- imported Ids first:
916 | not (isLocalId id) = fastBool (mayHaveCafRefs (idCafInfo id))
917 -- now Ids local to this module:
919 case lookupVarEnv p id of
920 Just id' -> fastBool (mayHaveCafRefs (idCafInfo id'))
921 Nothing -> fastBool False
923 cafRefs _ (Lit _) = fastBool False
924 cafRefs p (App f a) = fastOr (cafRefs p f) (cafRefs p) a
925 cafRefs p (Lam _ e) = cafRefs p e
926 cafRefs p (Let b e) = fastOr (cafRefss p (rhssOfBind b)) (cafRefs p) e
927 cafRefs p (Case e _bndr _ alts) = fastOr (cafRefs p e) (cafRefss p) (rhssOfAlts alts)
928 cafRefs p (Note _n e) = cafRefs p e
929 cafRefs p (Cast e _co) = cafRefs p e
930 cafRefs _ (Type _) = fastBool False
932 cafRefss :: VarEnv Id -> [Expr a] -> FastBool
933 cafRefss _ [] = fastBool False
934 cafRefss p (e:es) = fastOr (cafRefs p e) (cafRefss p) es
936 fastOr :: FastBool -> (a -> FastBool) -> a -> FastBool
937 -- hack for lazy-or over FastBool.
938 fastOr a f x = fastBool (isFastTrue a || isFastTrue (f x))