2 % (c) The University of Glasgow 2000
4 \section[ByteCodeGen]{Generate bytecode from Core}
7 module ByteCodeGen ( UnlinkedBCO, UnlinkedBCOExpr, ItblEnv, ClosureEnv, HValue,
9 byteCodeGen, coreExprToBCOs
12 #include "HsVersions.h"
15 import Name ( Name, getName )
16 import Id ( Id, idType, isDataConId_maybe, isPrimOpId_maybe, isFCallId,
17 idPrimRep, mkSysLocal, idName, isFCallId_maybe )
18 import ForeignCall ( ForeignCall(..), CCallTarget(..), CCallSpec(..) )
19 import OrdList ( OrdList, consOL, snocOL, appOL, unitOL,
20 nilOL, toOL, concatOL, fromOL )
21 import FiniteMap ( FiniteMap, addListToFM, listToFM,
22 addToFM, lookupFM, fmToList )
24 import PprCore ( pprCoreExpr )
25 import Literal ( Literal(..), literalPrimRep )
26 import PrimRep ( PrimRep(..) )
27 import PrimOp ( PrimOp(..) )
28 import CoreFVs ( freeVars )
29 import Type ( typePrimRep, splitTyConApp_maybe, isTyVarTy )
30 import DataCon ( dataConTag, fIRST_TAG, dataConTyCon,
31 dataConWrapId, isUnboxedTupleCon )
32 import TyCon ( TyCon(..), tyConFamilySize, isDataTyCon, tyConDataCons,
33 isFunTyCon, isUnboxedTupleTyCon )
34 import Class ( Class, classTyCon )
35 import Type ( Type, repType, splitRepFunTys )
36 import Util ( zipEqual, zipWith4Equal, naturalMergeSortLe, nOfThem )
37 import Var ( isTyVar )
38 import VarSet ( VarSet, varSetElems )
39 import PrimRep ( isFollowableRep )
40 import CmdLineOpts ( DynFlags, DynFlag(..) )
41 import ErrUtils ( showPass, dumpIfSet_dyn )
42 import Unique ( mkPseudoUnique3 )
43 import FastString ( FastString(..) )
44 import Panic ( GhcException(..) )
45 import PprType ( pprType )
46 import SMRep ( arrWordsHdrSize, arrPtrsHdrSize )
47 import Constants ( wORD_SIZE )
48 import ByteCodeInstr ( BCInstr(..), ProtoBCO(..), nameOfProtoBCO, bciStackUse )
49 import ByteCodeItbls ( ItblEnv, mkITbls )
50 import ByteCodeLink ( UnlinkedBCO, UnlinkedBCOExpr, assembleBCO,
51 ClosureEnv, HValue, filterNameMap,
52 iNTERP_STACK_CHECK_THRESH )
53 import ByteCodeFFI ( taggedSizeW, untaggedSizeW, mkMarshalCode )
54 import Linker ( lookupSymbol )
56 import List ( intersperse, sortBy, zip4 )
57 import Foreign ( Ptr(..), mallocBytes )
58 import Addr ( Addr(..), writeCharOffAddr )
59 import CTypes ( CInt )
60 import Exception ( throwDyn )
62 import PrelBase ( Int(..) )
63 import PrelGHC ( ByteArray# )
64 import PrelIOBase ( IO(..) )
69 %************************************************************************
71 \subsection{Functions visible from outside this module.}
73 %************************************************************************
77 byteCodeGen :: DynFlags
80 -> IO ([UnlinkedBCO], ItblEnv)
81 byteCodeGen dflags binds local_tycons local_classes
82 = do showPass dflags "ByteCodeGen"
83 let tycs = local_tycons ++ map classTyCon local_classes
84 itblenv <- mkITbls tycs
86 let flatBinds = concatMap getBind binds
87 getBind (NonRec bndr rhs) = [(bndr, freeVars rhs)]
88 getBind (Rec binds) = [(bndr, freeVars rhs) | (bndr,rhs) <- binds]
90 (BcM_State proto_bcos final_ctr mallocd, ())
91 <- runBc (BcM_State [] 0 [])
92 (mapBc (schemeR True) flatBinds `thenBc_` returnBc ())
94 when (not (null mallocd))
95 (panic "ByteCodeGen.byteCodeGen: missing final emitBc?")
97 dumpIfSet_dyn dflags Opt_D_dump_BCOs
98 "Proto-bcos" (vcat (intersperse (char ' ') (map ppr proto_bcos)))
100 bcos <- mapM assembleBCO proto_bcos
102 return (bcos, itblenv)
105 -- Returns: (the root BCO for this expression,
106 -- a list of auxilary BCOs resulting from compiling closures)
107 coreExprToBCOs :: DynFlags
109 -> IO UnlinkedBCOExpr
110 coreExprToBCOs dflags expr
111 = do showPass dflags "ByteCodeGen"
113 -- create a totally bogus name for the top-level BCO; this
114 -- should be harmless, since it's never used for anything
115 let invented_id = mkSysLocal SLIT("Expr-Top-Level") (mkPseudoUnique3 0)
116 (panic "invented_id's type")
117 let invented_name = idName invented_id
119 (BcM_State all_proto_bcos final_ctr mallocd, ())
120 <- runBc (BcM_State [] 0 [])
121 (schemeR True (invented_id, freeVars expr))
123 when (not (null mallocd))
124 (panic "ByteCodeGen.coreExprToBCOs: missing final emitBc?")
126 dumpIfSet_dyn dflags Opt_D_dump_BCOs
127 "Proto-bcos" (vcat (intersperse (char ' ') (map ppr all_proto_bcos)))
130 = case filter ((== invented_name).nameOfProtoBCO) all_proto_bcos of
131 [root_bco] -> root_bco
133 = filter ((/= invented_name).nameOfProtoBCO) all_proto_bcos
135 auxiliary_bcos <- mapM assembleBCO auxiliary_proto_bcos
136 root_bco <- assembleBCO root_proto_bco
138 return (root_bco, auxiliary_bcos)
141 %************************************************************************
143 \subsection{Compilation schema for the bytecode generator.}
145 %************************************************************************
149 type BCInstrList = OrdList BCInstr
151 type Sequel = Int -- back off to this depth before ENTER
153 -- Maps Ids to the offset from the stack _base_ so we don't have
154 -- to mess with it after each push/pop.
155 type BCEnv = FiniteMap Id Int -- To find vars on the stack
157 ppBCEnv :: BCEnv -> SDoc
160 $$ nest 4 (vcat (map pp_one (sortBy cmp_snd (fmToList p))))
163 pp_one (var, offset) = int offset <> colon <+> ppr var
164 cmp_snd x y = compare (snd x) (snd y)
166 -- Create a BCO and do a spot of peephole optimisation on the insns
168 mkProtoBCO nm instrs_ordlist origin mallocd_blocks
169 = ProtoBCO nm maybe_with_stack_check origin mallocd_blocks
171 -- Overestimate the stack usage (in words) of this BCO,
172 -- and if >= iNTERP_STACK_CHECK_THRESH, add an explicit
173 -- stack check. (The interpreter always does a stack check
174 -- for iNTERP_STACK_CHECK_THRESH words at the start of each
175 -- BCO anyway, so we only need to add an explicit on in the
176 -- (hopefully rare) cases when the (overestimated) stack use
177 -- exceeds iNTERP_STACK_CHECK_THRESH.
178 maybe_with_stack_check
179 | stack_overest >= 65535
180 = pprPanic "mkProtoBCO: stack use won't fit in 16 bits"
182 | stack_overest >= iNTERP_STACK_CHECK_THRESH
183 = (STKCHECK stack_overest) : peep_d
185 = peep_d -- the supposedly common case
187 stack_overest = sum (map bciStackUse peep_d)
188 + 10 {- just to be really really sure -}
191 -- Merge local pushes
192 peep_d = peep (fromOL instrs_ordlist)
194 peep (PUSH_L off1 : PUSH_L off2 : PUSH_L off3 : rest)
195 = PUSH_LLL off1 (off2-1) (off3-2) : peep rest
196 peep (PUSH_L off1 : PUSH_L off2 : rest)
197 = PUSH_LL off1 (off2-1) : peep rest
204 -- Compile code for the right hand side of a let binding.
205 -- Park the resulting BCO in the monad. Also requires the
206 -- variable to which this value was bound, so as to give the
207 -- resulting BCO a name. Bool indicates top-levelness.
209 schemeR :: Bool -> (Id, AnnExpr Id VarSet) -> BcM ()
210 schemeR is_top (nm, rhs)
214 $$ (ppr.filter (not.isTyVar).varSetElems.fst) rhs
215 $$ pprCoreExpr (deAnnotate rhs)
221 = schemeR_wrk is_top rhs nm (collect [] rhs)
224 collect xs (_, AnnNote note e)
226 collect xs (_, AnnLam x e)
227 = collect (if isTyVar x then xs else (x:xs)) e
228 collect xs not_lambda
229 = (reverse xs, not_lambda)
231 schemeR_wrk is_top original_body nm (args, body)
232 | Just dcon <- maybe_toplevel_null_con_rhs
233 = --trace ("nullary constructor! " ++ showSDocDebug (ppr nm)) (
234 emitBc (mkProtoBCO (getName nm) (toOL [PACK dcon 0, ENTER])
235 (Right original_body))
239 = let fvs = filter (not.isTyVar) (varSetElems (fst original_body))
240 all_args = reverse args ++ fvs
241 szsw_args = map taggedIdSizeW all_args
242 szw_args = sum szsw_args
243 p_init = listToFM (zip all_args (mkStackOffsets 0 szsw_args))
244 argcheck = unitOL (ARGCHECK szw_args)
246 schemeE szw_args 0 p_init body `thenBc` \ body_code ->
247 emitBc (mkProtoBCO (getName nm) (appOL argcheck body_code)
248 (Right original_body))
251 maybe_toplevel_null_con_rhs
252 | is_top && null args
255 -> case isDataConId_maybe v_wrk of
257 Just dc_wrk | nm == dataConWrapId dc_wrk
265 -- Let szsw be the sizes in words of some items pushed onto the stack,
266 -- which has initial depth d'. Return the values which the stack environment
267 -- should map these items to.
268 mkStackOffsets :: Int -> [Int] -> [Int]
269 mkStackOffsets original_depth szsw
270 = map (subtract 1) (tail (scanl (+) original_depth szsw))
272 -- Compile code to apply the given expression to the remaining args
273 -- on the stack, returning a HNF.
274 schemeE :: Int -> Sequel -> BCEnv -> AnnExpr Id VarSet -> BcM BCInstrList
276 -- Delegate tail-calls to schemeT.
277 schemeE d s p e@(fvs, AnnApp f a)
278 = schemeT d s p (fvs, AnnApp f a)
280 schemeE d s p e@(fvs, AnnVar v)
281 | isFollowableRep v_rep
282 = -- Ptr-ish thing; push it in the normal way
283 schemeT d s p (fvs, AnnVar v)
286 = -- returning an unboxed value. Heave it on the stack, SLIDE, and RETURN.
287 pushAtom True d p (AnnVar v) `thenBc` \ (push, szw) ->
288 returnBc (push -- value onto stack
289 `appOL` mkSLIDE szw (d-s) -- clear to sequel
290 `snocOL` RETURN v_rep) -- go
292 v_rep = typePrimRep (idType v)
294 schemeE d s p (fvs, AnnLit literal)
295 = pushAtom True d p (AnnLit literal) `thenBc` \ (push, szw) ->
296 let l_rep = literalPrimRep literal
297 in returnBc (push -- value onto stack
298 `appOL` mkSLIDE szw (d-s) -- clear to sequel
299 `snocOL` RETURN l_rep) -- go
301 schemeE d s p (fvs, AnnLet binds b)
302 = let (xs,rhss) = case binds of AnnNonRec x rhs -> ([x],[rhs])
303 AnnRec xs_n_rhss -> unzip xs_n_rhss
305 fvss = map (filter (not.isTyVar).varSetElems.fst) rhss
307 -- Sizes of tagged free vars, + 1 for the fn
308 sizes = map (\rhs_fvs -> 1 + sum (map taggedIdSizeW rhs_fvs)) fvss
310 -- This p', d' defn is safe because all the items being pushed
311 -- are ptrs, so all have size 1. d' and p' reflect the stack
312 -- after the closures have been allocated in the heap (but not
313 -- filled in), and pointers to them parked on the stack.
314 p' = addListToFM p (zipE xs (mkStackOffsets d (nOfThem n 1)))
317 infos = zipE4 fvss sizes xs [n, n-1 .. 1]
318 zipE = zipEqual "schemeE"
319 zipE4 = zipWith4Equal "schemeE" (\a b c d -> (a,b,c,d))
321 -- ToDo: don't build thunks for things with no free variables
322 buildThunk dd ([], size, id, off)
323 = returnBc (PUSH_G (Left (getName id))
324 `consOL` unitOL (MKAP (off+size-1) size))
325 buildThunk dd ((fv:fvs), size, id, off)
326 = pushAtom True dd p' (AnnVar fv)
327 `thenBc` \ (push_code, pushed_szw) ->
328 buildThunk (dd+pushed_szw) (fvs, size, id, off)
329 `thenBc` \ more_push_code ->
330 returnBc (push_code `appOL` more_push_code)
332 genThunkCode = mapBc (buildThunk d') infos `thenBc` \ tcodes ->
333 returnBc (concatOL tcodes)
335 allocCode = toOL (map ALLOC sizes)
337 schemeE d' s p' b `thenBc` \ bodyCode ->
338 mapBc (schemeR False) (zip xs rhss) `thenBc_`
339 genThunkCode `thenBc` \ thunkCode ->
340 returnBc (allocCode `appOL` thunkCode `appOL` bodyCode)
346 schemeE d s p (fvs_case, AnnCase (fvs_scrut, scrut) bndr
347 [(DEFAULT, [], (fvs_rhs, rhs))])
349 | let isFunType var_type
350 = case splitTyConApp_maybe var_type of
351 Just (tycon,_) | isFunTyCon tycon -> True
353 ty_bndr = repType (idType bndr)
354 in isFunType ty_bndr || isTyVarTy ty_bndr
357 -- case scrut::suspect of bndr { DEFAULT -> rhs }
359 -- let bndr = scrut in rhs
360 -- when suspect is polymorphic or arrowtyped
361 -- So the required strictness properties are not observed.
362 -- At some point, must fix this properly.
366 (AnnNonRec bndr (fvs_scrut, scrut)) (fvs_rhs, rhs)
369 in trace ("WARNING: ignoring polymorphic case in interpreted mode.\n" ++
370 " Possibly due to strict polymorphic/functional constructor args.\n" ++
371 " Your program may leak space unexpectedly.\n")
372 (schemeE d s p new_expr)
376 {- Convert case .... of (# VoidRep'd-thing, a #) -> ...
378 case .... of a -> ...
379 Use a as the name of the binder too.
381 Also case .... of (# a #) -> ...
383 case .... of a -> ...
385 schemeE d s p (fvs, AnnCase scrut bndr [(DataAlt dc, [bind1, bind2], rhs)])
386 | isUnboxedTupleCon dc && VoidRep == typePrimRep (idType bind1)
387 = --trace "automagic mashing of case alts (# VoidRep, a #)" (
388 schemeE d s p (fvs, AnnCase scrut bind2 [(DEFAULT, [bind2], rhs)])
391 schemeE d s p (fvs, AnnCase scrut bndr [(DataAlt dc, [bind1], rhs)])
392 | isUnboxedTupleCon dc
393 = --trace "automagic mashing of case alts (# a #)" (
394 schemeE d s p (fvs, AnnCase scrut bind1 [(DEFAULT, [bind1], rhs)])
397 schemeE d s p (fvs, AnnCase scrut bndr alts)
399 -- Top of stack is the return itbl, as usual.
400 -- underneath it is the pointer to the alt_code BCO.
401 -- When an alt is entered, it assumes the returned value is
402 -- on top of the itbl.
405 -- Env and depth in which to compile the alts, not including
406 -- any vars bound by the alts themselves
407 d' = d + ret_frame_sizeW + taggedIdSizeW bndr
408 p' = addToFM p bndr (d' - 1)
410 scrut_primrep = typePrimRep (idType bndr)
412 | scrut_primrep == PtrRep
414 | scrut_primrep `elem`
415 [CharRep, AddrRep, WordRep, IntRep, FloatRep, DoubleRep,
416 VoidRep, Int8Rep, Int16Rep, Int32Rep, Int64Rep,
417 Word8Rep, Word16Rep, Word32Rep, Word64Rep]
420 = pprPanic "ByteCodeGen.schemeE" (ppr scrut_primrep)
422 -- given an alt, return a discr and code for it.
423 codeAlt alt@(discr, binds_f, rhs)
425 = let (unpack_code, d_after_unpack, p_after_unpack)
426 = mkUnpackCode (filter (not.isTyVar) binds_f) d' p'
427 in schemeE d_after_unpack s p_after_unpack rhs
428 `thenBc` \ rhs_code ->
429 returnBc (my_discr alt, unpack_code `appOL` rhs_code)
431 = ASSERT(null binds_f)
432 schemeE d' s p' rhs `thenBc` \ rhs_code ->
433 returnBc (my_discr alt, rhs_code)
435 my_discr (DEFAULT, binds, rhs) = NoDiscr
436 my_discr (DataAlt dc, binds, rhs)
437 | isUnboxedTupleCon dc
438 = unboxedTupleException
440 = DiscrP (dataConTag dc - fIRST_TAG)
441 my_discr (LitAlt l, binds, rhs)
442 = case l of MachInt i -> DiscrI (fromInteger i)
443 MachFloat r -> DiscrF (fromRational r)
444 MachDouble r -> DiscrD (fromRational r)
445 MachChar i -> DiscrI i
446 _ -> pprPanic "schemeE(AnnCase).my_discr" (ppr l)
449 | not isAlgCase = Nothing
451 = case [dc | (DataAlt dc, _, _) <- alts] of
453 (dc:_) -> Just (tyConFamilySize (dataConTyCon dc))
456 mapBc codeAlt alts `thenBc` \ alt_stuff ->
457 mkMultiBranch maybe_ncons alt_stuff `thenBc` \ alt_final ->
459 alt_final_ac = ARGCHECK (taggedIdSizeW bndr) `consOL` alt_final
460 alt_bco_name = getName bndr
461 alt_bco = mkProtoBCO alt_bco_name alt_final_ac (Left alts)
463 schemeE (d + ret_frame_sizeW)
464 (d + ret_frame_sizeW) p scrut `thenBc` \ scrut_code ->
466 emitBc alt_bco `thenBc_`
467 returnBc (PUSH_AS alt_bco_name scrut_primrep `consOL` scrut_code)
470 schemeE d s p (fvs, AnnNote note body)
474 = pprPanic "ByteCodeGen.schemeE: unhandled case"
475 (pprCoreExpr (deAnnotate other))
478 -- Compile code to do a tail call. Specifically, push the fn,
479 -- slide the on-stack app back down to the sequel depth,
480 -- and enter. Four cases:
483 -- An application "PrelGHC.tagToEnum# <type> unboxed-int".
484 -- The int will be on the stack. Generate a code sequence
485 -- to convert it to the relevant constructor, SLIDE and ENTER.
487 -- 1. A nullary constructor. Push its closure on the stack
488 -- and SLIDE and RETURN.
490 -- 2. (Another nasty hack). Spot (# a::VoidRep, b #) and treat
491 -- it simply as b -- since the representations are identical
492 -- (the VoidRep takes up zero stack space). Also, spot
493 -- (# b #) and treat it as b.
495 -- 3. Application of a non-nullary constructor, by defn saturated.
496 -- Split the args into ptrs and non-ptrs, and push the nonptrs,
497 -- then the ptrs, and then do PACK and RETURN.
499 -- 4. Otherwise, it must be a function call. Push the args
500 -- right to left, SLIDE and ENTER.
502 schemeT :: Int -- Stack depth
503 -> Sequel -- Sequel depth
504 -> BCEnv -- stack env
510 -- | trace ("schemeT: env in = \n" ++ showSDocDebug (ppBCEnv p)) False
511 -- = panic "schemeT ?!?!"
513 -- | trace ("\nschemeT\n" ++ showSDoc (pprCoreExpr (deAnnotate app)) ++ "\n") False
517 | Just (arg, constr_names) <- maybe_is_tagToEnum_call
518 = pushAtom True d p arg `thenBc` \ (push, arg_words) ->
519 implement_tagToId constr_names `thenBc` \ tagToId_sequence ->
520 returnBc (push `appOL` tagToId_sequence
521 `appOL` mkSLIDE 1 (d+arg_words-s)
525 | is_con_call && null args_r_to_l
527 (PUSH_G (Left (getName con)) `consOL` mkSLIDE 1 (d-s))
532 | let isVoidRepAtom (_, AnnVar v) = VoidRep == typePrimRep (idType v)
533 isVoidRepAtom (_, AnnNote n e) = isVoidRepAtom e
534 in is_con_call && isUnboxedTupleCon con
535 && ( (length args_r_to_l == 2 && isVoidRepAtom (last (args_r_to_l)))
536 || (length args_r_to_l == 1)
538 = --trace (if length args_r_to_l == 1
539 -- then "schemeT: unboxed singleton"
540 -- else "schemeT: unboxed pair with Void first component") (
541 schemeT d s p (head args_r_to_l)
544 | Just (CCall ccall_spec) <- isFCallId_maybe fn
545 = generateCCall d s p ccall_spec fn args_r_to_l
549 = if is_con_call && isUnboxedTupleCon con
550 then unboxedTupleException
551 else do_pushery d (map snd args_final_r_to_l)
554 -- Detect and extract relevant info for the tagToEnum kludge.
555 maybe_is_tagToEnum_call
556 = let extract_constr_Names ty
557 = case splitTyConApp_maybe (repType ty) of
558 (Just (tyc, [])) | isDataTyCon tyc
559 -> map getName (tyConDataCons tyc)
560 other -> panic "maybe_is_tagToEnum_call.extract_constr_Ids"
563 (_, AnnApp (_, AnnApp (_, AnnVar v) (_, AnnType t)) arg)
564 -> case isPrimOpId_maybe v of
565 Just TagToEnumOp -> Just (snd arg, extract_constr_Names t)
569 -- Extract the args (R->L) and fn
570 (args_r_to_l_raw, fn) = chomp app
574 AnnApp f a -> case chomp f of (az, f) -> (a:az, f)
575 AnnNote n e -> chomp e
576 other -> pprPanic "schemeT"
577 (ppr (deAnnotate (panic "schemeT.chomp", other)))
579 args_r_to_l = filter (not.isTypeAtom.snd) args_r_to_l_raw
580 isTypeAtom (AnnType _) = True
583 -- decide if this is a constructor call, and rearrange
584 -- args appropriately.
585 maybe_dcon = isDataConId_maybe fn
586 is_con_call = case maybe_dcon of Nothing -> False; Just _ -> True
587 (Just con) = maybe_dcon
593 = filter (not.isPtr.snd) args_r_to_l ++ filter (isPtr.snd) args_r_to_l
594 where isPtr = isFollowableRep . atomRep
596 -- make code to push the args and then do the SLIDE-ENTER thing
597 tag_when_push = not is_con_call
598 narg_words = sum (map (get_arg_szw . atomRep . snd) args_r_to_l)
599 get_arg_szw = if tag_when_push then taggedSizeW else untaggedSizeW
601 do_pushery d (arg:args)
602 = pushAtom tag_when_push d p arg `thenBc` \ (push, arg_words) ->
603 do_pushery (d+arg_words) args `thenBc` \ more_push_code ->
604 returnBc (push `appOL` more_push_code)
606 | Just (CCall ccall_spec) <- isFCallId_maybe fn
607 = panic "schemeT.do_pushery: unexpected ccall"
610 Just con -> returnBc (
611 (PACK con narg_words `consOL`
612 mkSLIDE 1 (d - narg_words - s)) `snocOL`
616 -> pushAtom True d p (AnnVar fn)
617 `thenBc` \ (push, arg_words) ->
618 returnBc (push `appOL` mkSLIDE (narg_words+arg_words)
624 {- Deal with a CCall. Taggedly push the args onto the stack R->L,
625 deferencing ForeignObj#s and (ToDo: adjusting addrs to point to
626 payloads in Ptr/Byte arrays). Then, generate the marshalling
627 (machine) code for the ccall, and create bytecodes to call that and
628 then return in the right way.
630 generateCCall :: Int -> Sequel -- stack and sequel depths
632 -> CCallSpec -- where to call
633 -> Id -- of target, for type info
634 -> [AnnExpr Id VarSet] -- args (atoms)
637 generateCCall d0 s p ccall_spec@(CCallSpec target cconv safety) fn args_r_to_l
640 addr_usizeW = untaggedSizeW AddrRep
641 addr_tsizeW = taggedSizeW AddrRep
643 -- Get the args on the stack, with tags and suitably
644 -- dereferenced for the CCall. For each arg, return the
645 -- depth to the first word of the bits for that arg, and the
646 -- PrimRep of what was actually pushed.
648 pargs d [] = returnBc []
650 = let rep_arg = atomRep a
652 -- Don't push the FO; instead push the Addr# it
655 -> pushAtom False{-irrelevant-} d p a
656 `thenBc` \ (push_fo, _) ->
657 let foro_szW = taggedSizeW ForeignObjRep
658 d_now = d + addr_tsizeW
659 code = push_fo `appOL` toOL [
660 UPK_TAG addr_usizeW 0 0,
661 SLIDE addr_tsizeW foro_szW
663 in pargs d_now az `thenBc` \ rest ->
664 returnBc ((code, AddrRep) : rest)
667 -> pargs (d + addr_tsizeW) az `thenBc` \ rest ->
668 parg_ArrayishRep arrPtrsHdrSize d p a
670 returnBc ((code,AddrRep):rest)
673 -> pargs (d + addr_tsizeW) az `thenBc` \ rest ->
674 parg_ArrayishRep arrWordsHdrSize d p a
676 returnBc ((code,AddrRep):rest)
678 -- Default case: push taggedly, but otherwise intact.
680 -> pushAtom True d p a `thenBc` \ (code_a, sz_a) ->
681 pargs (d+sz_a) az `thenBc` \ rest ->
682 returnBc ((code_a, rep_arg) : rest)
684 -- Do magic for Ptr/Byte arrays. Push a ptr to the array on
685 -- the stack but then advance it over the headers, so as to
686 -- point to the payload.
687 parg_ArrayishRep hdrSizeW d p a
688 = pushAtom False{-irrel-} d p a `thenBc` \ (push_fo, _) ->
689 -- The ptr points at the header. Advance it over the
690 -- header and then pretend this is an Addr# (push a tag).
691 returnBc (push_fo `snocOL`
692 SWIZZLE 0 (hdrSizeW * untaggedSizeW PtrRep
695 PUSH_TAG addr_usizeW)
698 pargs d0 args_r_to_l `thenBc` \ code_n_reps ->
700 (pushs_arg, a_reps_pushed_r_to_l) = unzip code_n_reps
702 push_args = concatOL pushs_arg
703 d_after_args = d0 + sum (map taggedSizeW a_reps_pushed_r_to_l)
705 | null a_reps_pushed_r_to_l || head a_reps_pushed_r_to_l /= VoidRep
706 = panic "ByteCodeGen.generateCCall: missing or invalid World token?"
708 = reverse (tail a_reps_pushed_r_to_l)
710 -- Now: a_reps_pushed_RAW are the reps which are actually on the stack.
711 -- push_args is the code to do that.
712 -- d_after_args is the stack depth once the args are on.
714 -- Get the result rep.
715 (returns_void, r_rep)
716 = case maybe_getCCallReturnRep (idType fn) of
717 Nothing -> (True, VoidRep)
718 Just rr -> (False, rr)
720 Because the Haskell stack grows down, the a_reps refer to
721 lowest to highest addresses in that order. The args for the call
722 are on the stack. Now push an unboxed, tagged Addr# indicating
723 the C function to call. Then push a dummy placeholder for the
724 result. Finally, emit a CCALL insn with an offset pointing to the
725 Addr# just pushed, and a literal field holding the mallocville
726 address of the piece of marshalling code we generate.
727 So, just prior to the CCALL insn, the stack looks like this
728 (growing down, as usual):
733 Addr# address_of_C_fn
734 <placeholder-for-result#> (must be an unboxed type)
736 The interpreter then calls the marshall code mentioned
737 in the CCALL insn, passing it (& <placeholder-for-result#>),
738 that is, the addr of the topmost word in the stack.
739 When this returns, the placeholder will have been
740 filled in. The placeholder is slid down to the sequel
741 depth, and we RETURN.
743 This arrangement makes it simple to do f-i-dynamic since the Addr#
744 value is the first arg anyway. It also has the virtue that the
745 stack is GC-understandable at all times.
747 The marshalling code is generated specifically for this
748 call site, and so knows exactly the (Haskell) stack
749 offsets of the args, fn address and placeholder. It
750 copies the args to the C stack, calls the stacked addr,
751 and parks the result back in the placeholder. The interpreter
752 calls it as a normal C call, assuming it has a signature
753 void marshall_code ( StgWord* ptr_to_top_of_stack )
755 -- resolve static address
759 -> returnBc (False, panic "ByteCodeGen.generateCCall(dyn)")
761 -> ioToBc (lookupSymbol (_UNPK_ target)) `thenBc` \res ->
763 Just aa -> case aa of Ptr a# -> returnBc (True, A# a#)
764 Nothing -> returnBc invalid
768 invalid = pprPanic ("ByteCodeGen.generateCCall: unfindable "
769 ++ "symbol or otherwise invalid target")
772 get_target_info `thenBc` \ (is_static, static_target_addr) ->
775 -- Get the arg reps, zapping the leading Addr# in the dynamic case
776 a_reps -- | trace (showSDoc (ppr a_reps_pushed_RAW)) False = error "???"
777 | is_static = a_reps_pushed_RAW
778 | otherwise = if null a_reps_pushed_RAW
779 then panic "ByteCodeGen.generateCCall: dyn with no args"
780 else tail a_reps_pushed_RAW
783 (push_Addr, d_after_Addr)
785 = (toOL [PUSH_UBX (Right static_target_addr) addr_usizeW,
786 PUSH_TAG addr_usizeW],
787 d_after_args + addr_tsizeW)
788 | otherwise -- is already on the stack
789 = (nilOL, d_after_args)
791 -- Push the return placeholder. For a call returning nothing,
792 -- this is a VoidRep (tag).
793 r_usizeW = untaggedSizeW r_rep
794 r_tsizeW = taggedSizeW r_rep
795 d_after_r = d_after_Addr + r_tsizeW
796 r_lit = mkDummyLiteral r_rep
797 push_r = (if returns_void
799 else unitOL (PUSH_UBX (Left r_lit) r_usizeW))
801 unitOL (PUSH_TAG r_usizeW)
803 -- generate the marshalling code we're going to call
806 arg1_offW = r_tsizeW + addr_tsizeW
807 args_offW = map (arg1_offW +)
808 (init (scanl (+) 0 (map taggedSizeW a_reps)))
810 ioToBc (mkMarshalCode cconv
811 (r_offW, r_rep) addr_offW
812 (zip args_offW a_reps)) `thenBc` \ addr_of_marshaller ->
813 recordMallocBc addr_of_marshaller `thenBc_`
816 do_call = unitOL (CCALL addr_of_marshaller)
818 wrapup = mkSLIDE r_tsizeW (d_after_r - r_tsizeW - s)
819 `snocOL` RETURN r_rep
821 --trace (show (arg1_offW, args_offW , (map taggedSizeW a_reps) )) (
824 push_Addr `appOL` push_r `appOL` do_call `appOL` wrapup
829 -- Make a dummy literal, to be used as a placeholder for FFI return
830 -- values on the stack.
831 mkDummyLiteral :: PrimRep -> Literal
835 DoubleRep -> MachDouble 0
836 FloatRep -> MachFloat 0
837 AddrRep | taggedSizeW AddrRep == taggedSizeW WordRep -> MachWord 0
838 _ -> pprPanic "mkDummyLiteral" (ppr pr)
842 -- PrelGHC.Char# -> PrelGHC.State# PrelGHC.RealWorld
843 -- -> (# PrelGHC.State# PrelGHC.RealWorld, PrelGHC.Int# #)
846 -- and check that an unboxed pair is returned wherein the first arg is VoidRep'd.
848 -- Alternatively, for call-targets returning nothing, convert
850 -- PrelGHC.Char# -> PrelGHC.State# PrelGHC.RealWorld
851 -- -> (# PrelGHC.State# PrelGHC.RealWorld #)
855 maybe_getCCallReturnRep :: Type -> Maybe PrimRep
856 maybe_getCCallReturnRep fn_ty
857 = let (a_tys, r_ty) = splitRepFunTys fn_ty
859 = if length r_reps == 1 then Nothing else Just (r_reps !! 1)
861 = case splitTyConApp_maybe (repType r_ty) of
862 (Just (tyc, tys)) -> (tyc, map typePrimRep tys)
864 ok = ( (length r_reps == 2 && VoidRep == head r_reps)
865 || r_reps == [VoidRep] )
866 && isUnboxedTupleTyCon r_tycon
867 && case maybe_r_rep_to_go of
869 Just r_rep -> r_rep /= PtrRep
870 -- if it was, it would be impossible
871 -- to create a valid return value
872 -- placeholder on the stack
873 blargh = pprPanic "maybe_getCCallReturn: can't handle:"
876 --trace (showSDoc (ppr (a_reps, r_reps))) (
877 if ok then maybe_r_rep_to_go else blargh
880 atomRep (AnnVar v) = typePrimRep (idType v)
881 atomRep (AnnLit l) = literalPrimRep l
882 atomRep (AnnNote n b) = atomRep (snd b)
883 atomRep (AnnApp f (_, AnnType _)) = atomRep (snd f)
884 atomRep (AnnLam x e) | isTyVar x = atomRep (snd e)
885 atomRep other = pprPanic "atomRep" (ppr (deAnnotate (undefined,other)))
888 -- Compile code which expects an unboxed Int on the top of stack,
889 -- (call it i), and pushes the i'th closure in the supplied list
891 implement_tagToId :: [Name] -> BcM BCInstrList
892 implement_tagToId names
893 = ASSERT(not (null names))
894 getLabelsBc (length names) `thenBc` \ labels ->
895 getLabelBc `thenBc` \ label_fail ->
896 getLabelBc `thenBc` \ label_exit ->
897 zip4 labels (tail labels ++ [label_fail])
898 [0 ..] names `bind` \ infos ->
899 map (mkStep label_exit) infos `bind` \ steps ->
900 returnBc (concatOL steps
902 toOL [LABEL label_fail, CASEFAIL, LABEL label_exit])
904 mkStep l_exit (my_label, next_label, n, name_for_n)
905 = toOL [LABEL my_label,
906 TESTEQ_I n next_label,
907 PUSH_G (Left name_for_n),
911 -- Make code to unpack the top-of-stack constructor onto the stack,
912 -- adding tags for the unboxed bits. Takes the PrimReps of the
913 -- constructor's arguments. off_h and off_s are travelling offsets
914 -- along the constructor and the stack.
916 -- Supposing a constructor in the heap has layout
918 -- Itbl p_1 ... p_i np_1 ... np_j
920 -- then we add to the stack, shown growing down, the following:
932 -- so that in the common case (ptrs only) a single UNPACK instr can
933 -- copy all the payload of the constr onto the stack with no further ado.
935 mkUnpackCode :: [Id] -- constr args
936 -> Int -- depth before unpack
937 -> BCEnv -- env before unpack
938 -> (BCInstrList, Int, BCEnv)
939 mkUnpackCode vars d p
940 = --trace ("mkUnpackCode: " ++ showSDocDebug (ppr vars)
941 -- ++ " --> " ++ show d' ++ "\n" ++ showSDocDebug (ppBCEnv p')
943 (code_p `appOL` code_np, d', p')
947 vreps = [(var, typePrimRep (idType var)) | var <- vars]
949 -- ptrs and nonptrs, forward
950 vreps_p = filter (isFollowableRep.snd) vreps
951 vreps_np = filter (not.isFollowableRep.snd) vreps
953 -- the order in which we will augment the environment
954 vreps_env = reverse vreps_p ++ reverse vreps_np
957 vreps_env_tszsw = map (taggedSizeW.snd) vreps_env
958 p' = addListToFM p (zip (map fst vreps_env)
959 (mkStackOffsets d vreps_env_tszsw))
960 d' = d + sum vreps_env_tszsw
962 -- code to unpack the ptrs
963 ptrs_szw = sum (map (untaggedSizeW.snd) vreps_p)
964 code_p | null vreps_p = nilOL
965 | otherwise = unitOL (UNPACK ptrs_szw)
967 -- code to unpack the nonptrs
968 vreps_env_uszw = sum (map (untaggedSizeW.snd) vreps_env)
969 code_np = do_nptrs vreps_env_uszw ptrs_szw (reverse (map snd vreps_np))
970 do_nptrs off_h off_s [] = nilOL
971 do_nptrs off_h off_s (npr:nprs)
972 | npr `elem` [IntRep, WordRep, FloatRep, DoubleRep, CharRep, AddrRep]
975 = pprPanic "ByteCodeGen.mkUnpackCode" (ppr npr)
977 approved = UPK_TAG usizeW (off_h-usizeW) off_s `consOL` theRest
978 theRest = do_nptrs (off_h-usizeW) (off_s + tsizeW) nprs
979 usizeW = untaggedSizeW npr
980 tsizeW = taggedSizeW npr
983 -- Push an atom onto the stack, returning suitable code & number of
984 -- stack words used. Pushes it either tagged or untagged, since
985 -- pushAtom is used to set up the stack prior to copying into the
986 -- heap for both APs (requiring tags) and constructors (which don't).
988 -- NB this means NO GC between pushing atoms for a constructor and
989 -- copying them into the heap. It probably also means that
990 -- tail calls MUST be of the form atom{atom ... atom} since if the
991 -- expression head was allowed to be arbitrary, there could be GC
992 -- in between pushing the arg atoms and completing the head.
993 -- (not sure; perhaps the allocate/doYouWantToGC interface means this
994 -- isn't a problem; but only if arbitrary graph construction for the
995 -- head doesn't leave this BCO, since GC might happen at the start of
996 -- each BCO (we consult doYouWantToGC there).
998 -- Blargh. JRS 001206
1000 -- NB (further) that the env p must map each variable to the highest-
1001 -- numbered stack slot for it. For example, if the stack has depth 4
1002 -- and we tagged-ly push (v :: Int#) on it, the value will be in stack[4],
1003 -- the tag in stack[5], the stack will have depth 6, and p must map v to
1004 -- 5 and not to 4. Stack locations are numbered from zero, so a depth
1005 -- 6 stack has valid words 0 .. 5.
1007 pushAtom :: Bool -> Int -> BCEnv -> AnnExpr' Id VarSet -> BcM (BCInstrList, Int)
1008 pushAtom tagged d p (AnnVar v)
1010 | idPrimRep v == VoidRep
1011 = if tagged then returnBc (unitOL (PUSH_TAG 0), 1)
1012 else panic "ByteCodeGen.pushAtom(VoidRep,untaggedly)"
1015 = pprPanic "pushAtom: shouldn't get an FCallId here" (ppr v)
1017 | Just primop <- isPrimOpId_maybe v
1018 = returnBc (unitOL (PUSH_G (Right primop)), 1)
1022 str = "\npushAtom " ++ showSDocDebug (ppr v)
1023 ++ " :: " ++ showSDocDebug (pprType (idType v))
1024 ++ ", depth = " ++ show d
1025 ++ ", tagged = " ++ show tagged ++ ", env =\n" ++
1026 showSDocDebug (ppBCEnv p)
1027 ++ " --> words: " ++ show (snd result) ++ "\n" ++
1028 showSDoc (nest 4 (vcat (map ppr (fromOL (fst result)))))
1029 ++ "\nendPushAtom " ++ showSDocDebug (ppr v)
1033 = case lookupBCEnv_maybe p v of
1034 Just d_v -> (toOL (nOfThem nwords (PUSH_L (d-d_v+sz_t-2))), nwords)
1035 Nothing -> ASSERT(sz_t == 1) (unitOL (PUSH_G (Left nm)), nwords)
1037 nm = case isDataConId_maybe v of
1039 Nothing -> getName v
1041 sz_t = taggedIdSizeW v
1042 sz_u = untaggedIdSizeW v
1043 nwords = if tagged then sz_t else sz_u
1047 pushAtom True d p (AnnLit lit)
1048 = pushAtom False d p (AnnLit lit) `thenBc` \ (ubx_code, ubx_size) ->
1049 returnBc (ubx_code `snocOL` PUSH_TAG ubx_size, 1 + ubx_size)
1051 pushAtom False d p (AnnLit lit)
1053 MachWord w -> code WordRep
1054 MachInt i -> code IntRep
1055 MachFloat r -> code FloatRep
1056 MachDouble r -> code DoubleRep
1057 MachChar c -> code CharRep
1058 MachStr s -> pushStr s
1061 = let size_host_words = untaggedSizeW rep
1062 in returnBc (unitOL (PUSH_UBX (Left lit) size_host_words),
1066 = let getMallocvilleAddr
1068 CharStr s i -> returnBc (A# s)
1070 FastString _ l ba ->
1071 -- sigh, a string in the heap is no good to us.
1072 -- We need a static C pointer, since the type of
1073 -- a string literal is Addr#. So, copy the string
1074 -- into C land and introduce a memory leak
1075 -- at the same time.
1077 -- CAREFUL! Chars are 32 bits in ghc 4.09+
1078 in ioToBc (mallocBytes (n+1)) `thenBc` \ (Ptr a#) ->
1079 recordMallocBc (A# a#) `thenBc_`
1081 do strncpy (Ptr a#) ba (fromIntegral n)
1082 writeCharOffAddr (A# a#) n '\0'
1085 other -> panic "ByteCodeGen.pushAtom.pushStr"
1087 getMallocvilleAddr `thenBc` \ addr ->
1088 -- Get the addr on the stack, untaggedly
1089 returnBc (unitOL (PUSH_UBX (Right addr) 1), 1)
1095 pushAtom tagged d p (AnnApp f (_, AnnType _))
1096 = pushAtom tagged d p (snd f)
1098 pushAtom tagged d p (AnnNote note e)
1099 = pushAtom tagged d p (snd e)
1101 pushAtom tagged d p (AnnLam x e)
1103 = pushAtom tagged d p (snd e)
1105 pushAtom tagged d p other
1106 = pprPanic "ByteCodeGen.pushAtom"
1107 (pprCoreExpr (deAnnotate (undefined, other)))
1109 foreign import "strncpy" strncpy :: Ptr a -> ByteArray# -> CInt -> IO ()
1112 -- Given a bunch of alts code and their discrs, do the donkey work
1113 -- of making a multiway branch using a switch tree.
1114 -- What a load of hassle!
1115 mkMultiBranch :: Maybe Int -- # datacons in tycon, if alg alt
1116 -- a hint; generates better code
1117 -- Nothing is always safe
1118 -> [(Discr, BCInstrList)]
1120 mkMultiBranch maybe_ncons raw_ways
1121 = let d_way = filter (isNoDiscr.fst) raw_ways
1122 notd_ways = naturalMergeSortLe
1123 (\w1 w2 -> leAlt (fst w1) (fst w2))
1124 (filter (not.isNoDiscr.fst) raw_ways)
1126 mkTree :: [(Discr, BCInstrList)] -> Discr -> Discr -> BcM BCInstrList
1127 mkTree [] range_lo range_hi = returnBc the_default
1129 mkTree [val] range_lo range_hi
1130 | range_lo `eqAlt` range_hi
1131 = returnBc (snd val)
1133 = getLabelBc `thenBc` \ label_neq ->
1134 returnBc (mkTestEQ (fst val) label_neq
1136 `appOL` unitOL (LABEL label_neq)
1137 `appOL` the_default))
1139 mkTree vals range_lo range_hi
1140 = let n = length vals `div` 2
1141 vals_lo = take n vals
1142 vals_hi = drop n vals
1143 v_mid = fst (head vals_hi)
1145 getLabelBc `thenBc` \ label_geq ->
1146 mkTree vals_lo range_lo (dec v_mid) `thenBc` \ code_lo ->
1147 mkTree vals_hi v_mid range_hi `thenBc` \ code_hi ->
1148 returnBc (mkTestLT v_mid label_geq
1150 `appOL` unitOL (LABEL label_geq)
1154 = case d_way of [] -> unitOL CASEFAIL
1157 -- None of these will be needed if there are no non-default alts
1158 (mkTestLT, mkTestEQ, init_lo, init_hi)
1160 = panic "mkMultiBranch: awesome foursome"
1162 = case fst (head notd_ways) of {
1163 DiscrI _ -> ( \(DiscrI i) fail_label -> TESTLT_I i fail_label,
1164 \(DiscrI i) fail_label -> TESTEQ_I i fail_label,
1167 DiscrF _ -> ( \(DiscrF f) fail_label -> TESTLT_F f fail_label,
1168 \(DiscrF f) fail_label -> TESTEQ_F f fail_label,
1171 DiscrD _ -> ( \(DiscrD d) fail_label -> TESTLT_D d fail_label,
1172 \(DiscrD d) fail_label -> TESTEQ_D d fail_label,
1175 DiscrP _ -> ( \(DiscrP i) fail_label -> TESTLT_P i fail_label,
1176 \(DiscrP i) fail_label -> TESTEQ_P i fail_label,
1178 DiscrP algMaxBound )
1181 (algMinBound, algMaxBound)
1182 = case maybe_ncons of
1183 Just n -> (0, n - 1)
1184 Nothing -> (minBound, maxBound)
1186 (DiscrI i1) `eqAlt` (DiscrI i2) = i1 == i2
1187 (DiscrF f1) `eqAlt` (DiscrF f2) = f1 == f2
1188 (DiscrD d1) `eqAlt` (DiscrD d2) = d1 == d2
1189 (DiscrP i1) `eqAlt` (DiscrP i2) = i1 == i2
1190 NoDiscr `eqAlt` NoDiscr = True
1193 (DiscrI i1) `leAlt` (DiscrI i2) = i1 <= i2
1194 (DiscrF f1) `leAlt` (DiscrF f2) = f1 <= f2
1195 (DiscrD d1) `leAlt` (DiscrD d2) = d1 <= d2
1196 (DiscrP i1) `leAlt` (DiscrP i2) = i1 <= i2
1197 NoDiscr `leAlt` NoDiscr = True
1200 isNoDiscr NoDiscr = True
1203 dec (DiscrI i) = DiscrI (i-1)
1204 dec (DiscrP i) = DiscrP (i-1)
1205 dec other = other -- not really right, but if you
1206 -- do cases on floating values, you'll get what you deserve
1208 -- same snotty comment applies to the following
1210 minD, maxD :: Double
1216 mkTree notd_ways init_lo init_hi
1220 %************************************************************************
1222 \subsection{Supporting junk for the compilation schemes}
1224 %************************************************************************
1228 -- Describes case alts
1236 instance Outputable Discr where
1237 ppr (DiscrI i) = int i
1238 ppr (DiscrF f) = text (show f)
1239 ppr (DiscrD d) = text (show d)
1240 ppr (DiscrP i) = int i
1241 ppr NoDiscr = text "DEF"
1244 -- Find things in the BCEnv (the what's-on-the-stack-env)
1245 -- See comment preceding pushAtom for precise meaning of env contents
1246 --lookupBCEnv :: BCEnv -> Id -> Int
1247 --lookupBCEnv env nm
1248 -- = case lookupFM env nm of
1249 -- Nothing -> pprPanic "lookupBCEnv"
1250 -- (ppr nm $$ char ' ' $$ vcat (map ppr (fmToList env)))
1253 lookupBCEnv_maybe :: BCEnv -> Id -> Maybe Int
1254 lookupBCEnv_maybe = lookupFM
1257 taggedIdSizeW, untaggedIdSizeW :: Id -> Int
1258 taggedIdSizeW = taggedSizeW . typePrimRep . idType
1259 untaggedIdSizeW = untaggedSizeW . typePrimRep . idType
1261 unboxedTupleException :: a
1262 unboxedTupleException
1265 ("Bytecode generator can't handle unboxed tuples. Possibly due\n" ++
1266 "\tto foreign import/export decls in source. Workaround:\n" ++
1267 "\tcompile this module to a .o file, then restart session."))
1270 mkSLIDE n d = if d == 0 then nilOL else unitOL (SLIDE n d)
1275 %************************************************************************
1277 \subsection{The bytecode generator's monad}
1279 %************************************************************************
1283 = BcM_State { bcos :: [ProtoBCO Name], -- accumulates completed BCOs
1284 nextlabel :: Int, -- for generating local labels
1285 malloced :: [Addr] } -- ptrs malloced for current BCO
1286 -- Should be free()d when it is GCd
1287 type BcM r = BcM_State -> IO (BcM_State, r)
1289 ioToBc :: IO a -> BcM a
1290 ioToBc io st = do x <- io
1293 runBc :: BcM_State -> BcM r -> IO (BcM_State, r)
1294 runBc st0 m = do (st1, res) <- m st0
1297 thenBc :: BcM a -> (a -> BcM b) -> BcM b
1298 thenBc expr cont st0
1299 = do (st1, q) <- expr st0
1300 (st2, r) <- cont q st1
1303 thenBc_ :: BcM a -> BcM b -> BcM b
1304 thenBc_ expr cont st0
1305 = do (st1, q) <- expr st0
1306 (st2, r) <- cont st1
1309 returnBc :: a -> BcM a
1310 returnBc result st = return (st, result)
1313 mapBc :: (a -> BcM b) -> [a] -> BcM [b]
1314 mapBc f [] = returnBc []
1316 = f x `thenBc` \ r ->
1317 mapBc f xs `thenBc` \ rs ->
1320 emitBc :: ([Addr] -> ProtoBCO Name) -> BcM ()
1322 = return (st{bcos = bco (malloced st) : bcos st, malloced=[]}, ())
1326 | not (null (malloced st))
1327 = panic "ByteCodeGen.newbcoBc: missed prior emitBc?"
1331 recordMallocBc :: Addr -> BcM ()
1333 = return (st{malloced = a : malloced st}, ())
1335 getLabelBc :: BcM Int
1337 = return (st{nextlabel = 1 + nextlabel st}, nextlabel st)
1339 getLabelsBc :: Int -> BcM [Int]
1341 = let ctr = nextlabel st
1342 in return (st{nextlabel = ctr+n}, [ctr .. ctr+n-1])