2 % (c) The University of Glasgow 2002-2006
5 ByteCodeGen: Generate bytecode from Core
8 module ByteCodeGen ( UnlinkedBCO, byteCodeGen, coreExprToBCOs ) where
10 #include "HsVersions.h"
50 import Data.List ( intersperse, sortBy, zip4, zip6, partition )
51 import Foreign ( Ptr, castPtr, mallocBytes, pokeByteOff, Word8,
52 withForeignPtr, castFunPtrToPtr, nullPtr, plusPtr )
54 import Control.Exception ( throwDyn )
56 import GHC.Exts ( Int(..), ByteArray# )
58 import Control.Monad ( when )
59 import Data.Char ( ord, chr )
67 -- -----------------------------------------------------------------------------
68 -- Generating byte code for a complete module
70 byteCodeGen :: DynFlags
74 -> IO CompiledByteCode
75 byteCodeGen dflags binds tycs modBreaks
76 = do showPass dflags "ByteCodeGen"
78 let flatBinds = [ (bndr, freeVars rhs)
79 | (bndr, rhs) <- flattenBinds binds]
81 us <- mkSplitUniqSupply 'y'
82 (BcM_State _us final_ctr mallocd _, proto_bcos)
83 <- runBc us modBreaks (mapM schemeTopBind flatBinds)
85 when (notNull mallocd)
86 (panic "ByteCodeGen.byteCodeGen: missing final emitBc?")
88 dumpIfSet_dyn dflags Opt_D_dump_BCOs
89 "Proto-BCOs" (vcat (intersperse (char ' ') (map ppr proto_bcos)))
91 assembleBCOs proto_bcos tycs
93 -- -----------------------------------------------------------------------------
94 -- Generating byte code for an expression
96 -- Returns: (the root BCO for this expression,
97 -- a list of auxilary BCOs resulting from compiling closures)
98 coreExprToBCOs :: DynFlags
101 coreExprToBCOs dflags expr
102 = do showPass dflags "ByteCodeGen"
104 -- create a totally bogus name for the top-level BCO; this
105 -- should be harmless, since it's never used for anything
106 let invented_name = mkSystemVarName (mkPseudoUniqueE 0) FSLIT("ExprTopLevel")
107 invented_id = Id.mkLocalId invented_name (panic "invented_id's type")
109 -- the uniques are needed to generate fresh variables when we introduce new
110 -- let bindings for ticked expressions
111 us <- mkSplitUniqSupply 'y'
112 (BcM_State _us final_ctr mallocd _ , proto_bco)
113 <- runBc us emptyModBreaks (schemeTopBind (invented_id, freeVars expr))
115 when (notNull mallocd)
116 (panic "ByteCodeGen.coreExprToBCOs: missing final emitBc?")
118 dumpIfSet_dyn dflags Opt_D_dump_BCOs "Proto-BCOs" (ppr proto_bco)
120 assembleBCO proto_bco
123 -- -----------------------------------------------------------------------------
124 -- Compilation schema for the bytecode generator
126 type BCInstrList = OrdList BCInstr
128 type Sequel = Int -- back off to this depth before ENTER
130 -- Maps Ids to the offset from the stack _base_ so we don't have
131 -- to mess with it after each push/pop.
132 type BCEnv = FiniteMap Id Int -- To find vars on the stack
134 ppBCEnv :: BCEnv -> SDoc
137 $$ nest 4 (vcat (map pp_one (sortBy cmp_snd (fmToList p))))
140 pp_one (var, offset) = int offset <> colon <+> ppr var <+> ppr (idCgRep var)
141 cmp_snd x y = compare (snd x) (snd y)
143 -- Create a BCO and do a spot of peephole optimisation on the insns
148 -> Either [AnnAlt Id VarSet] (AnnExpr Id VarSet)
152 -> Bool -- True <=> is a return point, rather than a function
155 mkProtoBCO nm instrs_ordlist origin arity bitmap_size bitmap is_ret mallocd_blocks
158 protoBCOInstrs = maybe_with_stack_check,
159 protoBCOBitmap = bitmap,
160 protoBCOBitmapSize = bitmap_size,
161 protoBCOArity = arity,
162 protoBCOExpr = origin,
163 protoBCOPtrs = mallocd_blocks
166 -- Overestimate the stack usage (in words) of this BCO,
167 -- and if >= iNTERP_STACK_CHECK_THRESH, add an explicit
168 -- stack check. (The interpreter always does a stack check
169 -- for iNTERP_STACK_CHECK_THRESH words at the start of each
170 -- BCO anyway, so we only need to add an explicit on in the
171 -- (hopefully rare) cases when the (overestimated) stack use
172 -- exceeds iNTERP_STACK_CHECK_THRESH.
173 maybe_with_stack_check
175 -- don't do stack checks at return points;
176 -- everything is aggregated up to the top BCO
177 -- (which must be a function)
178 | stack_overest >= iNTERP_STACK_CHECK_THRESH
179 = STKCHECK stack_overest : peep_d
181 = peep_d -- the supposedly common case
183 -- We assume that this sum doesn't wrap
184 stack_overest = sum (map bciStackUse peep_d)
186 -- Merge local pushes
187 peep_d = peep (fromOL instrs_ordlist)
189 peep (PUSH_L off1 : PUSH_L off2 : PUSH_L off3 : rest)
190 = PUSH_LLL off1 (off2-1) (off3-2) : peep rest
191 peep (PUSH_L off1 : PUSH_L off2 : rest)
192 = PUSH_LL off1 (off2-1) : peep rest
198 argBits :: [CgRep] -> [Bool]
201 | isFollowableArg rep = False : argBits args
202 | otherwise = take (cgRepSizeW rep) (repeat True) ++ argBits args
204 -- -----------------------------------------------------------------------------
207 -- Compile code for the right-hand side of a top-level binding
209 schemeTopBind :: (Id, AnnExpr Id VarSet) -> BcM (ProtoBCO Name)
212 schemeTopBind (id, rhs)
213 | Just data_con <- isDataConWorkId_maybe id,
214 isNullaryRepDataCon data_con = do
215 -- Special case for the worker of a nullary data con.
216 -- It'll look like this: Nil = /\a -> Nil a
217 -- If we feed it into schemeR, we'll get
219 -- because mkConAppCode treats nullary constructor applications
220 -- by just re-using the single top-level definition. So
221 -- for the worker itself, we must allocate it directly.
222 -- ioToBc (putStrLn $ "top level BCO")
223 emitBc (mkProtoBCO (getName id) (toOL [PACK data_con 0, ENTER])
224 (Right rhs) 0 0 [{-no bitmap-}] False{-not alts-})
227 = schemeR [{- No free variables -}] (id, rhs)
230 -- -----------------------------------------------------------------------------
233 -- Compile code for a right-hand side, to give a BCO that,
234 -- when executed with the free variables and arguments on top of the stack,
235 -- will return with a pointer to the result on top of the stack, after
236 -- removing the free variables and arguments.
238 -- Park the resulting BCO in the monad. Also requires the
239 -- variable to which this value was bound, so as to give the
240 -- resulting BCO a name.
242 schemeR :: [Id] -- Free vars of the RHS, ordered as they
243 -- will appear in the thunk. Empty for
244 -- top-level things, which have no free vars.
245 -> (Id, AnnExpr Id VarSet)
246 -> BcM (ProtoBCO Name)
247 schemeR fvs (nm, rhs)
251 $$ (ppr.filter (not.isTyVar).varSetElems.fst) rhs
252 $$ pprCoreExpr (deAnnotate rhs)
258 = schemeR_wrk fvs nm rhs (collect [] rhs)
260 collect :: [Var] -> AnnExpr Id VarSet -> ([Var], AnnExpr' Id VarSet)
261 collect xs (_, AnnNote note e) = collect xs e
262 collect xs (_, AnnCast e _) = collect xs e
263 collect xs (_, AnnLam x e) = collect (if isTyVar x then xs else (x:xs)) e
264 collect xs (_, not_lambda) = (reverse xs, not_lambda)
266 schemeR_wrk :: [Id] -> Id -> AnnExpr Id VarSet -> ([Var], AnnExpr' Var VarSet) -> BcM (ProtoBCO Name)
267 schemeR_wrk fvs nm original_body (args, body)
269 all_args = reverse args ++ fvs
270 arity = length all_args
271 -- all_args are the args in reverse order. We're compiling a function
272 -- \fv1..fvn x1..xn -> e
273 -- i.e. the fvs come first
275 szsw_args = map idSizeW all_args
276 szw_args = sum szsw_args
277 p_init = listToFM (zip all_args (mkStackOffsets 0 szsw_args))
279 -- make the arg bitmap
280 bits = argBits (reverse (map idCgRep all_args))
281 bitmap_size = length bits
282 bitmap = mkBitmap bits
284 body_code <- schemeER_wrk szw_args p_init body
286 emitBc (mkProtoBCO (getName nm) body_code (Right original_body)
287 arity bitmap_size bitmap False{-not alts-})
289 -- introduce break instructions for ticked expressions
290 schemeER_wrk :: Int -> BCEnv -> AnnExpr' Id VarSet -> BcM BCInstrList
292 | Just (tickInfo, (_annot, newRhs)) <- isTickedExp' rhs = do
293 code <- schemeE d 0 p newRhs
295 let idOffSets = getVarOffSets d p tickInfo
296 let tickNumber = tickInfo_number tickInfo
297 let breakInfo = BreakInfo
298 { breakInfo_module = tickInfo_module tickInfo
299 , breakInfo_number = tickNumber
300 , breakInfo_vars = idOffSets
301 , breakInfo_resty = exprType (deAnnotate' newRhs)
303 let breakInstr = case arr of (BA arr#) -> BRK_FUN arr# tickNumber breakInfo
304 return $ breakInstr `consOL` code
305 | otherwise = schemeE d 0 p rhs
307 getVarOffSets :: Int -> BCEnv -> TickInfo -> [(Id, Int)]
308 getVarOffSets d p = catMaybes . map (getOffSet d p) . tickInfo_locals
310 getOffSet :: Int -> BCEnv -> Id -> Maybe (Id, Int)
312 = case lookupBCEnv_maybe env id of
314 Just offset -> Just (id, d - offset)
316 fvsToEnv :: BCEnv -> VarSet -> [Id]
317 -- Takes the free variables of a right-hand side, and
318 -- delivers an ordered list of the local variables that will
319 -- be captured in the thunk for the RHS
320 -- The BCEnv argument tells which variables are in the local
321 -- environment: these are the ones that should be captured
323 -- The code that constructs the thunk, and the code that executes
324 -- it, have to agree about this layout
325 fvsToEnv p fvs = [v | v <- varSetElems fvs,
326 isId v, -- Could be a type variable
329 -- -----------------------------------------------------------------------------
334 { tickInfo_number :: Int -- the (module) unique number of the tick
335 , tickInfo_module :: Module -- the origin of the ticked expression
336 , tickInfo_locals :: [Id] -- the local vars in scope at the ticked expression
339 instance Outputable TickInfo where
340 ppr info = text "TickInfo" <+>
341 parens (int (tickInfo_number info) <+> ppr (tickInfo_module info) <+>
342 ppr (tickInfo_locals info))
344 -- Compile code to apply the given expression to the remaining args
345 -- on the stack, returning a HNF.
346 schemeE :: Int -> Sequel -> BCEnv -> AnnExpr' Id VarSet -> BcM BCInstrList
348 -- Delegate tail-calls to schemeT.
349 schemeE d s p e@(AnnApp f a)
352 schemeE d s p e@(AnnVar v)
353 | not (isUnLiftedType v_type)
354 = -- Lifted-type thing; push it in the normal way
358 = do -- Returning an unlifted value.
359 -- Heave it on the stack, SLIDE, and RETURN.
360 (push, szw) <- pushAtom d p (AnnVar v)
361 return (push -- value onto stack
362 `appOL` mkSLIDE szw (d-s) -- clear to sequel
363 `snocOL` RETURN_UBX v_rep) -- go
366 v_rep = typeCgRep v_type
368 schemeE d s p (AnnLit literal)
369 = do (push, szw) <- pushAtom d p (AnnLit literal)
370 let l_rep = typeCgRep (literalType literal)
371 return (push -- value onto stack
372 `appOL` mkSLIDE szw (d-s) -- clear to sequel
373 `snocOL` RETURN_UBX l_rep) -- go
375 schemeE d s p (AnnLet (AnnNonRec x (_,rhs)) (_,body))
376 | (AnnVar v, args_r_to_l) <- splitApp rhs,
377 Just data_con <- isDataConWorkId_maybe v,
378 dataConRepArity data_con == length args_r_to_l
379 = do -- Special case for a non-recursive let whose RHS is a
380 -- saturatred constructor application.
381 -- Just allocate the constructor and carry on
382 alloc_code <- mkConAppCode d s p data_con args_r_to_l
383 body_code <- schemeE (d+1) s (addToFM p x d) body
384 return (alloc_code `appOL` body_code)
386 -- General case for let. Generates correct, if inefficient, code in
388 schemeE d s p (AnnLet binds (_,body))
389 = let (xs,rhss) = case binds of AnnNonRec x rhs -> ([x],[rhs])
390 AnnRec xs_n_rhss -> unzip xs_n_rhss
393 fvss = map (fvsToEnv p' . fst) rhss
395 -- Sizes of free vars
396 sizes = map (\rhs_fvs -> sum (map idSizeW rhs_fvs)) fvss
398 -- the arity of each rhs
399 arities = map (length . fst . collect []) rhss
401 -- This p', d' defn is safe because all the items being pushed
402 -- are ptrs, so all have size 1. d' and p' reflect the stack
403 -- after the closures have been allocated in the heap (but not
404 -- filled in), and pointers to them parked on the stack.
405 p' = addListToFM p (zipE xs (mkStackOffsets d (nOfThem n_binds 1)))
407 zipE = zipEqual "schemeE"
409 -- ToDo: don't build thunks for things with no free variables
410 build_thunk dd [] size bco off arity
411 = return (PUSH_BCO bco `consOL` unitOL (mkap (off+size) size))
413 mkap | arity == 0 = MKAP
415 build_thunk dd (fv:fvs) size bco off arity = do
416 (push_code, pushed_szw) <- pushAtom dd p' (AnnVar fv)
417 more_push_code <- build_thunk (dd+pushed_szw) fvs size bco off arity
418 return (push_code `appOL` more_push_code)
420 alloc_code = toOL (zipWith mkAlloc sizes arities)
421 where mkAlloc sz 0 = ALLOC_AP sz
422 mkAlloc sz arity = ALLOC_PAP arity sz
424 compile_bind d' fvs x rhs size arity off = do
425 bco <- schemeR fvs (x,rhs)
426 build_thunk d' fvs size bco off arity
429 [ compile_bind d' fvs x rhs size arity n
430 | (fvs, x, rhs, size, arity, n) <-
431 zip6 fvss xs rhss sizes arities [n_binds, n_binds-1 .. 1]
434 body_code <- schemeE d' s p' body
435 thunk_codes <- sequence compile_binds
436 return (alloc_code `appOL` concatOL thunk_codes `appOL` body_code)
438 -- introduce a let binding for a ticked case expression. This rule
439 -- *should* only fire when the expression was not already let-bound
440 -- (the code gen for let bindings should take care of that). Todo: we
441 -- call exprFreeVars on a deAnnotated expression, this may not be the
442 -- best way to calculate the free vars but it seemed like the least
443 -- intrusive thing to do
444 schemeE d s p exp@(AnnCase {})
445 | Just (tickInfo,rhs) <- isTickedExp' exp
446 = if isUnLiftedType ty
447 then schemeE d s p (snd rhs)
450 -- Todo: is emptyVarSet correct on the next line?
451 let letExp = AnnLet (AnnNonRec id (fvs, exp)) (emptyVarSet, AnnVar id)
453 where exp' = deAnnotate' exp
454 fvs = exprFreeVars exp'
457 schemeE d s p (AnnCase scrut bndr _ [(DataAlt dc, [bind1, bind2], rhs)])
458 | isUnboxedTupleCon dc, VoidArg <- typeCgRep (idType bind1)
460 -- case .... of x { (# VoidArg'd-thing, a #) -> ... }
462 -- case .... of a { DEFAULT -> ... }
463 -- becuse the return convention for both are identical.
465 -- Note that it does not matter losing the void-rep thing from the
466 -- envt (it won't be bound now) because we never look such things up.
468 = --trace "automagic mashing of case alts (# VoidArg, a #)" $
469 doCase d s p scrut bind2 [(DEFAULT, [], rhs)] True{-unboxed tuple-}
471 | isUnboxedTupleCon dc, VoidArg <- typeCgRep (idType bind2)
472 = --trace "automagic mashing of case alts (# a, VoidArg #)" $
473 doCase d s p scrut bind1 [(DEFAULT, [], rhs)] True{-unboxed tuple-}
475 schemeE d s p (AnnCase scrut bndr _ [(DataAlt dc, [bind1], rhs)])
476 | isUnboxedTupleCon dc
477 -- Similarly, convert
478 -- case .... of x { (# a #) -> ... }
480 -- case .... of a { DEFAULT -> ... }
481 = --trace "automagic mashing of case alts (# a #)" $
482 doCase d s p scrut bind1 [(DEFAULT, [], rhs)] True{-unboxed tuple-}
484 schemeE d s p (AnnCase scrut bndr _ alts)
485 = doCase d s p scrut bndr alts False{-not an unboxed tuple-}
487 schemeE d s p (AnnNote note (_, body))
490 schemeE d s p (AnnCast (_, body) _)
494 = pprPanic "ByteCodeGen.schemeE: unhandled case"
495 (pprCoreExpr (deAnnotate' other))
501 A ticked expression looks like this:
503 case tick<n> var1 ... varN of DEFAULT -> e
505 (*) <n> is the number of the tick, which is unique within a module
506 (*) var1 ... varN are the local variables in scope at the tick site
508 If we find a ticked expression we return:
510 Just ((n, [var1 ... varN]), e)
512 otherwise we return Nothing.
514 The idea is that the "case tick<n> ..." is really just an annotation on
515 the code. When we find such a thing, we pull out the useful information,
516 and then compile the code as if it was just the expression "e".
520 isTickedExp :: AnnExpr Id a -> Maybe (TickInfo, AnnExpr Id a)
521 isTickedExp (annot, expr) = isTickedExp' expr
523 isTickedExp' :: AnnExpr' Id a -> Maybe (TickInfo, AnnExpr Id a)
524 isTickedExp' (AnnCase scrut _bndr _type alts)
525 | Just tickInfo <- isTickedScrut scrut,
526 [(DEFAULT, _bndr, rhs)] <- alts
527 = Just (tickInfo, rhs)
529 isTickedScrut :: (AnnExpr Id a) -> Maybe TickInfo
532 Just (TickBox modName tickNumber) <- isTickBoxOp_maybe id
533 = Just $ TickInfo { tickInfo_number = tickNumber
534 , tickInfo_module = modName
535 , tickInfo_locals = idsOfArgs args
537 | otherwise = Nothing
539 (f, args) = collectArgs $ deAnnotate expr
540 idsOfArgs :: [Expr Id] -> [Id]
541 idsOfArgs = catMaybes . map exprId
542 exprId :: Expr Id -> Maybe Id
543 exprId (Var id) = Just id
544 exprId other = Nothing
546 isTickedExp' other = Nothing
548 -- Compile code to do a tail call. Specifically, push the fn,
549 -- slide the on-stack app back down to the sequel depth,
550 -- and enter. Four cases:
553 -- An application "GHC.Prim.tagToEnum# <type> unboxed-int".
554 -- The int will be on the stack. Generate a code sequence
555 -- to convert it to the relevant constructor, SLIDE and ENTER.
557 -- 1. The fn denotes a ccall. Defer to generateCCall.
559 -- 2. (Another nasty hack). Spot (# a::VoidArg, b #) and treat
560 -- it simply as b -- since the representations are identical
561 -- (the VoidArg takes up zero stack space). Also, spot
562 -- (# b #) and treat it as b.
564 -- 3. Application of a constructor, by defn saturated.
565 -- Split the args into ptrs and non-ptrs, and push the nonptrs,
566 -- then the ptrs, and then do PACK and RETURN.
568 -- 4. Otherwise, it must be a function call. Push the args
569 -- right to left, SLIDE and ENTER.
571 schemeT :: Int -- Stack depth
572 -> Sequel -- Sequel depth
573 -> BCEnv -- stack env
574 -> AnnExpr' Id VarSet
579 -- | trace ("schemeT: env in = \n" ++ showSDocDebug (ppBCEnv p)) False
580 -- = panic "schemeT ?!?!"
582 -- | trace ("\nschemeT\n" ++ showSDoc (pprCoreExpr (deAnnotate' app)) ++ "\n") False
586 | Just (arg, constr_names) <- maybe_is_tagToEnum_call
587 = do (push, arg_words) <- pushAtom d p arg
588 tagToId_sequence <- implement_tagToId constr_names
589 return (push `appOL` tagToId_sequence
590 `appOL` mkSLIDE 1 (d+arg_words-s)
594 | Just (CCall ccall_spec) <- isFCallId_maybe fn
595 = generateCCall d s p ccall_spec fn args_r_to_l
597 -- Case 2: Constructor application
598 | Just con <- maybe_saturated_dcon,
599 isUnboxedTupleCon con
600 = case args_r_to_l of
601 [arg1,arg2] | isVoidArgAtom arg1 ->
602 unboxedTupleReturn d s p arg2
603 [arg1,arg2] | isVoidArgAtom arg2 ->
604 unboxedTupleReturn d s p arg1
605 _other -> unboxedTupleException
607 -- Case 3: Ordinary data constructor
608 | Just con <- maybe_saturated_dcon
609 = do alloc_con <- mkConAppCode d s p con args_r_to_l
610 return (alloc_con `appOL`
611 mkSLIDE 1 (d - s) `snocOL`
614 -- Case 4: Tail call of function
616 = doTailCall d s p fn args_r_to_l
619 -- Detect and extract relevant info for the tagToEnum kludge.
620 maybe_is_tagToEnum_call
621 = let extract_constr_Names ty
622 | Just (tyc, []) <- splitTyConApp_maybe (repType ty),
624 = map (getName . dataConWorkId) (tyConDataCons tyc)
625 -- NOTE: use the worker name, not the source name of
626 -- the DataCon. See DataCon.lhs for details.
628 = panic "maybe_is_tagToEnum_call.extract_constr_Ids"
631 (AnnApp (_, AnnApp (_, AnnVar v) (_, AnnType t)) arg)
632 -> case isPrimOpId_maybe v of
633 Just TagToEnumOp -> Just (snd arg, extract_constr_Names t)
637 -- Extract the args (R->L) and fn
638 -- The function will necessarily be a variable,
639 -- because we are compiling a tail call
640 (AnnVar fn, args_r_to_l) = splitApp app
642 -- Only consider this to be a constructor application iff it is
643 -- saturated. Otherwise, we'll call the constructor wrapper.
644 n_args = length args_r_to_l
646 = case isDataConWorkId_maybe fn of
647 Just con | dataConRepArity con == n_args -> Just con
650 -- -----------------------------------------------------------------------------
651 -- Generate code to build a constructor application,
652 -- leaving it on top of the stack
654 mkConAppCode :: Int -> Sequel -> BCEnv
655 -> DataCon -- The data constructor
656 -> [AnnExpr' Id VarSet] -- Args, in *reverse* order
659 mkConAppCode orig_d s p con [] -- Nullary constructor
660 = ASSERT( isNullaryRepDataCon con )
661 return (unitOL (PUSH_G (getName (dataConWorkId con))))
662 -- Instead of doing a PACK, which would allocate a fresh
663 -- copy of this constructor, use the single shared version.
665 mkConAppCode orig_d s p con args_r_to_l
666 = ASSERT( dataConRepArity con == length args_r_to_l )
667 do_pushery orig_d (non_ptr_args ++ ptr_args)
669 -- The args are already in reverse order, which is the way PACK
670 -- expects them to be. We must push the non-ptrs after the ptrs.
671 (ptr_args, non_ptr_args) = partition isPtrAtom args_r_to_l
673 do_pushery d (arg:args)
674 = do (push, arg_words) <- pushAtom d p arg
675 more_push_code <- do_pushery (d+arg_words) args
676 return (push `appOL` more_push_code)
678 = return (unitOL (PACK con n_arg_words))
680 n_arg_words = d - orig_d
683 -- -----------------------------------------------------------------------------
684 -- Returning an unboxed tuple with one non-void component (the only
685 -- case we can handle).
687 -- Remember, we don't want to *evaluate* the component that is being
688 -- returned, even if it is a pointed type. We always just return.
691 :: Int -> Sequel -> BCEnv
692 -> AnnExpr' Id VarSet -> BcM BCInstrList
693 unboxedTupleReturn d s p arg = do
694 (push, sz) <- pushAtom d p arg
696 mkSLIDE sz (d-s) `snocOL`
697 RETURN_UBX (atomRep arg))
699 -- -----------------------------------------------------------------------------
700 -- Generate code for a tail-call
703 :: Int -> Sequel -> BCEnv
704 -> Id -> [AnnExpr' Id VarSet]
706 doTailCall init_d s p fn args
707 = do_pushes init_d args (map atomRep args)
709 do_pushes d [] reps = do
710 ASSERT( null reps ) return ()
711 (push_fn, sz) <- pushAtom d p (AnnVar fn)
712 ASSERT( sz == 1 ) return ()
713 return (push_fn `appOL` (
714 mkSLIDE ((d-init_d) + 1) (init_d - s) `appOL`
716 do_pushes d args reps = do
717 let (push_apply, n, rest_of_reps) = findPushSeq reps
718 (these_args, rest_of_args) = splitAt n args
719 (next_d, push_code) <- push_seq d these_args
720 instrs <- do_pushes (next_d + 1) rest_of_args rest_of_reps
721 -- ^^^ for the PUSH_APPLY_ instruction
722 return (push_code `appOL` (push_apply `consOL` instrs))
724 push_seq d [] = return (d, nilOL)
725 push_seq d (arg:args) = do
726 (push_code, sz) <- pushAtom d p arg
727 (final_d, more_push_code) <- push_seq (d+sz) args
728 return (final_d, push_code `appOL` more_push_code)
730 -- v. similar to CgStackery.findMatch, ToDo: merge
731 findPushSeq (PtrArg: PtrArg: PtrArg: PtrArg: PtrArg: PtrArg: rest)
732 = (PUSH_APPLY_PPPPPP, 6, rest)
733 findPushSeq (PtrArg: PtrArg: PtrArg: PtrArg: PtrArg: rest)
734 = (PUSH_APPLY_PPPPP, 5, rest)
735 findPushSeq (PtrArg: PtrArg: PtrArg: PtrArg: rest)
736 = (PUSH_APPLY_PPPP, 4, rest)
737 findPushSeq (PtrArg: PtrArg: PtrArg: rest)
738 = (PUSH_APPLY_PPP, 3, rest)
739 findPushSeq (PtrArg: PtrArg: rest)
740 = (PUSH_APPLY_PP, 2, rest)
741 findPushSeq (PtrArg: rest)
742 = (PUSH_APPLY_P, 1, rest)
743 findPushSeq (VoidArg: rest)
744 = (PUSH_APPLY_V, 1, rest)
745 findPushSeq (NonPtrArg: rest)
746 = (PUSH_APPLY_N, 1, rest)
747 findPushSeq (FloatArg: rest)
748 = (PUSH_APPLY_F, 1, rest)
749 findPushSeq (DoubleArg: rest)
750 = (PUSH_APPLY_D, 1, rest)
751 findPushSeq (LongArg: rest)
752 = (PUSH_APPLY_L, 1, rest)
754 = panic "ByteCodeGen.findPushSeq"
756 -- -----------------------------------------------------------------------------
759 doCase :: Int -> Sequel -> BCEnv
760 -> AnnExpr Id VarSet -> Id -> [AnnAlt Id VarSet]
761 -> Bool -- True <=> is an unboxed tuple case, don't enter the result
763 doCase d s p (_,scrut) bndr alts is_unboxed_tuple
765 -- Top of stack is the return itbl, as usual.
766 -- underneath it is the pointer to the alt_code BCO.
767 -- When an alt is entered, it assumes the returned value is
768 -- on top of the itbl.
771 -- An unlifted value gets an extra info table pushed on top
772 -- when it is returned.
773 unlifted_itbl_sizeW | isAlgCase = 0
776 -- depth of stack after the return value has been pushed
777 d_bndr = d + ret_frame_sizeW + idSizeW bndr
779 -- depth of stack after the extra info table for an unboxed return
780 -- has been pushed, if any. This is the stack depth at the
782 d_alts = d_bndr + unlifted_itbl_sizeW
784 -- Env in which to compile the alts, not including
785 -- any vars bound by the alts themselves
786 p_alts = addToFM p bndr (d_bndr - 1)
788 bndr_ty = idType bndr
789 isAlgCase = not (isUnLiftedType bndr_ty) && not is_unboxed_tuple
791 -- given an alt, return a discr and code for it.
792 codeAlt alt@(DEFAULT, _, (_,rhs))
793 = do rhs_code <- schemeE d_alts s p_alts rhs
794 return (NoDiscr, rhs_code)
796 codeAlt alt@(discr, bndrs, (_,rhs))
797 -- primitive or nullary constructor alt: no need to UNPACK
798 | null real_bndrs = do
799 rhs_code <- schemeE d_alts s p_alts rhs
800 return (my_discr alt, rhs_code)
801 -- algebraic alt with some binders
802 | ASSERT(isAlgCase) otherwise =
804 (ptrs,nptrs) = partition (isFollowableArg.idCgRep) real_bndrs
805 ptr_sizes = map idSizeW ptrs
806 nptrs_sizes = map idSizeW nptrs
807 bind_sizes = ptr_sizes ++ nptrs_sizes
808 size = sum ptr_sizes + sum nptrs_sizes
809 -- the UNPACK instruction unpacks in reverse order...
810 p' = addListToFM p_alts
811 (zip (reverse (ptrs ++ nptrs))
812 (mkStackOffsets d_alts (reverse bind_sizes)))
814 rhs_code <- schemeE (d_alts+size) s p' rhs
815 return (my_discr alt, unitOL (UNPACK size) `appOL` rhs_code)
817 real_bndrs = filter (not.isTyVar) bndrs
819 my_discr (DEFAULT, binds, rhs) = NoDiscr {-shouldn't really happen-}
820 my_discr (DataAlt dc, binds, rhs)
821 | isUnboxedTupleCon dc
822 = unboxedTupleException
824 = DiscrP (dataConTag dc - fIRST_TAG)
825 my_discr (LitAlt l, binds, rhs)
826 = case l of MachInt i -> DiscrI (fromInteger i)
827 MachFloat r -> DiscrF (fromRational r)
828 MachDouble r -> DiscrD (fromRational r)
829 MachChar i -> DiscrI (ord i)
830 _ -> pprPanic "schemeE(AnnCase).my_discr" (ppr l)
833 | not isAlgCase = Nothing
835 = case [dc | (DataAlt dc, _, _) <- alts] of
837 (dc:_) -> Just (tyConFamilySize (dataConTyCon dc))
839 -- the bitmap is relative to stack depth d, i.e. before the
840 -- BCO, info table and return value are pushed on.
841 -- This bit of code is v. similar to buildLivenessMask in CgBindery,
842 -- except that here we build the bitmap from the known bindings of
843 -- things that are pointers, whereas in CgBindery the code builds the
844 -- bitmap from the free slots and unboxed bindings.
847 -- NOTE [7/12/2006] bug #1013, testcase ghci/should_run/ghci002.
848 -- The bitmap must cover the portion of the stack up to the sequel only.
849 -- Previously we were building a bitmap for the whole depth (d), but we
850 -- really want a bitmap up to depth (d-s). This affects compilation of
851 -- case-of-case expressions, which is the only time we can be compiling a
852 -- case expression with s /= 0.
854 bitmap = intsToReverseBitmap bitmap_size{-size-}
855 (sortLe (<=) (filter (< bitmap_size) rel_slots))
858 rel_slots = concat (map spread binds)
860 | isFollowableArg (idCgRep id) = [ rel_offset ]
862 where rel_offset = d - offset - 1
865 alt_stuff <- mapM codeAlt alts
866 alt_final <- mkMultiBranch maybe_ncons alt_stuff
869 alt_bco_name = getName bndr
870 alt_bco = mkProtoBCO alt_bco_name alt_final (Left alts)
871 0{-no arity-} bitmap_size bitmap True{-is alts-}
873 -- trace ("case: bndr = " ++ showSDocDebug (ppr bndr) ++ "\ndepth = " ++ show d ++ "\nenv = \n" ++ showSDocDebug (ppBCEnv p) ++
874 -- "\n bitmap = " ++ show bitmap) $ do
875 scrut_code <- schemeE (d + ret_frame_sizeW) (d + ret_frame_sizeW) p scrut
876 alt_bco' <- emitBc alt_bco
878 | isAlgCase = PUSH_ALTS alt_bco'
879 | otherwise = PUSH_ALTS_UNLIFTED alt_bco' (typeCgRep bndr_ty)
880 return (push_alts `consOL` scrut_code)
883 -- -----------------------------------------------------------------------------
884 -- Deal with a CCall.
886 -- Taggedly push the args onto the stack R->L,
887 -- deferencing ForeignObj#s and adjusting addrs to point to
888 -- payloads in Ptr/Byte arrays. Then, generate the marshalling
889 -- (machine) code for the ccall, and create bytecodes to call that and
890 -- then return in the right way.
892 generateCCall :: Int -> Sequel -- stack and sequel depths
894 -> CCallSpec -- where to call
895 -> Id -- of target, for type info
896 -> [AnnExpr' Id VarSet] -- args (atoms)
899 generateCCall d0 s p ccall_spec@(CCallSpec target cconv safety) fn args_r_to_l
902 addr_sizeW = cgRepSizeW NonPtrArg
904 -- Get the args on the stack, with tags and suitably
905 -- dereferenced for the CCall. For each arg, return the
906 -- depth to the first word of the bits for that arg, and the
907 -- CgRep of what was actually pushed.
909 pargs d [] = return []
911 = let arg_ty = repType (exprType (deAnnotate' a))
913 in case splitTyConApp_maybe arg_ty of
914 -- Don't push the FO; instead push the Addr# it
917 | t == arrayPrimTyCon || t == mutableArrayPrimTyCon
918 -> do rest <- pargs (d + addr_sizeW) az
919 code <- parg_ArrayishRep arrPtrsHdrSize d p a
920 return ((code,NonPtrArg):rest)
922 | t == byteArrayPrimTyCon || t == mutableByteArrayPrimTyCon
923 -> do rest <- pargs (d + addr_sizeW) az
924 code <- parg_ArrayishRep arrWordsHdrSize d p a
925 return ((code,NonPtrArg):rest)
927 -- Default case: push taggedly, but otherwise intact.
929 -> do (code_a, sz_a) <- pushAtom d p a
930 rest <- pargs (d+sz_a) az
931 return ((code_a, atomRep a) : rest)
933 -- Do magic for Ptr/Byte arrays. Push a ptr to the array on
934 -- the stack but then advance it over the headers, so as to
935 -- point to the payload.
936 parg_ArrayishRep hdrSize d p a
937 = do (push_fo, _) <- pushAtom d p a
938 -- The ptr points at the header. Advance it over the
939 -- header and then pretend this is an Addr#.
940 return (push_fo `snocOL` SWIZZLE 0 hdrSize)
943 code_n_reps <- pargs d0 args_r_to_l
945 (pushs_arg, a_reps_pushed_r_to_l) = unzip code_n_reps
947 push_args = concatOL pushs_arg
948 d_after_args = d0 + sum (map cgRepSizeW a_reps_pushed_r_to_l)
950 | null a_reps_pushed_r_to_l || head a_reps_pushed_r_to_l /= VoidArg
951 = panic "ByteCodeGen.generateCCall: missing or invalid World token?"
953 = reverse (tail a_reps_pushed_r_to_l)
955 -- Now: a_reps_pushed_RAW are the reps which are actually on the stack.
956 -- push_args is the code to do that.
957 -- d_after_args is the stack depth once the args are on.
959 -- Get the result rep.
960 (returns_void, r_rep)
961 = case maybe_getCCallReturnRep (idType fn) of
962 Nothing -> (True, VoidArg)
963 Just rr -> (False, rr)
965 Because the Haskell stack grows down, the a_reps refer to
966 lowest to highest addresses in that order. The args for the call
967 are on the stack. Now push an unboxed Addr# indicating
968 the C function to call. Then push a dummy placeholder for the
969 result. Finally, emit a CCALL insn with an offset pointing to the
970 Addr# just pushed, and a literal field holding the mallocville
971 address of the piece of marshalling code we generate.
972 So, just prior to the CCALL insn, the stack looks like this
973 (growing down, as usual):
978 Addr# address_of_C_fn
979 <placeholder-for-result#> (must be an unboxed type)
981 The interpreter then calls the marshall code mentioned
982 in the CCALL insn, passing it (& <placeholder-for-result#>),
983 that is, the addr of the topmost word in the stack.
984 When this returns, the placeholder will have been
985 filled in. The placeholder is slid down to the sequel
986 depth, and we RETURN.
988 This arrangement makes it simple to do f-i-dynamic since the Addr#
989 value is the first arg anyway.
991 The marshalling code is generated specifically for this
992 call site, and so knows exactly the (Haskell) stack
993 offsets of the args, fn address and placeholder. It
994 copies the args to the C stack, calls the stacked addr,
995 and parks the result back in the placeholder. The interpreter
996 calls it as a normal C call, assuming it has a signature
997 void marshall_code ( StgWord* ptr_to_top_of_stack )
999 -- resolve static address
1003 -> return (False, panic "ByteCodeGen.generateCCall(dyn)")
1005 -> do res <- ioToBc (lookupStaticPtr target)
1008 (is_static, static_target_addr) <- get_target_info
1011 -- Get the arg reps, zapping the leading Addr# in the dynamic case
1012 a_reps -- | trace (showSDoc (ppr a_reps_pushed_RAW)) False = error "???"
1013 | is_static = a_reps_pushed_RAW
1014 | otherwise = if null a_reps_pushed_RAW
1015 then panic "ByteCodeGen.generateCCall: dyn with no args"
1016 else tail a_reps_pushed_RAW
1019 (push_Addr, d_after_Addr)
1021 = (toOL [PUSH_UBX (Right static_target_addr) addr_sizeW],
1022 d_after_args + addr_sizeW)
1023 | otherwise -- is already on the stack
1024 = (nilOL, d_after_args)
1026 -- Push the return placeholder. For a call returning nothing,
1027 -- this is a VoidArg (tag).
1028 r_sizeW = cgRepSizeW r_rep
1029 d_after_r = d_after_Addr + r_sizeW
1030 r_lit = mkDummyLiteral r_rep
1031 push_r = (if returns_void
1033 else unitOL (PUSH_UBX (Left r_lit) r_sizeW))
1035 -- generate the marshalling code we're going to call
1038 arg1_offW = r_sizeW + addr_sizeW
1039 args_offW = map (arg1_offW +)
1040 (init (scanl (+) 0 (map cgRepSizeW a_reps)))
1042 addr_of_marshaller <- ioToBc (mkMarshalCode cconv
1043 (r_offW, r_rep) addr_offW
1044 (zip args_offW a_reps))
1045 recordItblMallocBc (ItblPtr (castFunPtrToPtr addr_of_marshaller))
1047 -- Offset of the next stack frame down the stack. The CCALL
1048 -- instruction needs to describe the chunk of stack containing
1049 -- the ccall args to the GC, so it needs to know how large it
1050 -- is. See comment in Interpreter.c with the CCALL instruction.
1051 stk_offset = d_after_r - s
1054 do_call = unitOL (CCALL stk_offset (castFunPtrToPtr addr_of_marshaller))
1056 wrapup = mkSLIDE r_sizeW (d_after_r - r_sizeW - s)
1057 `snocOL` RETURN_UBX r_rep
1059 --trace (show (arg1_offW, args_offW , (map cgRepSizeW a_reps) )) $
1062 push_Addr `appOL` push_r `appOL` do_call `appOL` wrapup
1066 -- Make a dummy literal, to be used as a placeholder for FFI return
1067 -- values on the stack.
1068 mkDummyLiteral :: CgRep -> Literal
1071 NonPtrArg -> MachWord 0
1072 DoubleArg -> MachDouble 0
1073 FloatArg -> MachFloat 0
1074 LongArg -> MachWord64 0
1075 _ -> moan64 "mkDummyLiteral" (ppr pr)
1079 -- GHC.Prim.Char# -> GHC.Prim.State# GHC.Prim.RealWorld
1080 -- -> (# GHC.Prim.State# GHC.Prim.RealWorld, GHC.Prim.Int# #)
1083 -- and check that an unboxed pair is returned wherein the first arg is VoidArg'd.
1085 -- Alternatively, for call-targets returning nothing, convert
1087 -- GHC.Prim.Char# -> GHC.Prim.State# GHC.Prim.RealWorld
1088 -- -> (# GHC.Prim.State# GHC.Prim.RealWorld #)
1092 maybe_getCCallReturnRep :: Type -> Maybe CgRep
1093 maybe_getCCallReturnRep fn_ty
1094 = let (a_tys, r_ty) = splitFunTys (dropForAlls fn_ty)
1096 = if isSingleton r_reps then Nothing else Just (r_reps !! 1)
1098 = case splitTyConApp_maybe (repType r_ty) of
1099 (Just (tyc, tys)) -> (tyc, map typeCgRep tys)
1101 ok = ( ( r_reps `lengthIs` 2 && VoidArg == head r_reps)
1102 || r_reps == [VoidArg] )
1103 && isUnboxedTupleTyCon r_tycon
1104 && case maybe_r_rep_to_go of
1106 Just r_rep -> r_rep /= PtrArg
1107 -- if it was, it would be impossible
1108 -- to create a valid return value
1109 -- placeholder on the stack
1110 blargh = pprPanic "maybe_getCCallReturn: can't handle:"
1113 --trace (showSDoc (ppr (a_reps, r_reps))) $
1114 if ok then maybe_r_rep_to_go else blargh
1116 -- Compile code which expects an unboxed Int on the top of stack,
1117 -- (call it i), and pushes the i'th closure in the supplied list
1118 -- as a consequence.
1119 implement_tagToId :: [Name] -> BcM BCInstrList
1120 implement_tagToId names
1121 = ASSERT( notNull names )
1122 do labels <- getLabelsBc (length names)
1123 label_fail <- getLabelBc
1124 label_exit <- getLabelBc
1125 let infos = zip4 labels (tail labels ++ [label_fail])
1127 steps = map (mkStep label_exit) infos
1128 return (concatOL steps
1130 toOL [LABEL label_fail, CASEFAIL, LABEL label_exit])
1132 mkStep l_exit (my_label, next_label, n, name_for_n)
1133 = toOL [LABEL my_label,
1134 TESTEQ_I n next_label,
1139 -- -----------------------------------------------------------------------------
1142 -- Push an atom onto the stack, returning suitable code & number of
1143 -- stack words used.
1145 -- The env p must map each variable to the highest- numbered stack
1146 -- slot for it. For example, if the stack has depth 4 and we
1147 -- tagged-ly push (v :: Int#) on it, the value will be in stack[4],
1148 -- the tag in stack[5], the stack will have depth 6, and p must map v
1149 -- to 5 and not to 4. Stack locations are numbered from zero, so a
1150 -- depth 6 stack has valid words 0 .. 5.
1152 pushAtom :: Int -> BCEnv -> AnnExpr' Id VarSet -> BcM (BCInstrList, Int)
1154 pushAtom d p (AnnApp f (_, AnnType _))
1155 = pushAtom d p (snd f)
1157 pushAtom d p (AnnNote note e)
1158 = pushAtom d p (snd e)
1160 pushAtom d p (AnnLam x e)
1162 = pushAtom d p (snd e)
1164 pushAtom d p (AnnVar v)
1166 | idCgRep v == VoidArg
1170 = pprPanic "pushAtom: shouldn't get an FCallId here" (ppr v)
1172 | Just primop <- isPrimOpId_maybe v
1173 = return (unitOL (PUSH_PRIMOP primop), 1)
1175 | Just d_v <- lookupBCEnv_maybe p v -- v is a local variable
1176 = return (toOL (nOfThem sz (PUSH_L (d-d_v+sz-2))), sz)
1177 -- d - d_v the number of words between the TOS
1178 -- and the 1st slot of the object
1180 -- d - d_v - 1 the offset from the TOS of the 1st slot
1182 -- d - d_v - 1 + sz - 1 the offset from the TOS of the last slot
1185 -- Having found the last slot, we proceed to copy the right number of
1186 -- slots on to the top of the stack.
1188 | otherwise -- v must be a global variable
1190 return (unitOL (PUSH_G (getName v)), sz)
1196 pushAtom d p (AnnLit lit)
1198 MachLabel fs _ -> code NonPtrArg
1199 MachWord w -> code NonPtrArg
1200 MachInt i -> code PtrArg
1201 MachFloat r -> code FloatArg
1202 MachDouble r -> code DoubleArg
1203 MachChar c -> code NonPtrArg
1204 MachStr s -> pushStr s
1207 = let size_host_words = cgRepSizeW rep
1208 in return (unitOL (PUSH_UBX (Left lit) size_host_words),
1212 = let getMallocvilleAddr
1214 FastString _ n _ fp _ ->
1215 -- we could grab the Ptr from the ForeignPtr,
1216 -- but then we have no way to control its lifetime.
1217 -- In reality it'll probably stay alive long enoungh
1218 -- by virtue of the global FastString table, but
1219 -- to be on the safe side we copy the string into
1220 -- a malloc'd area of memory.
1221 do ptr <- ioToBc (mallocBytes (n+1))
1224 withForeignPtr fp $ \p -> do
1225 memcpy ptr p (fromIntegral n)
1226 pokeByteOff ptr n (fromIntegral (ord '\0') :: Word8)
1230 addr <- getMallocvilleAddr
1231 -- Get the addr on the stack, untaggedly
1232 return (unitOL (PUSH_UBX (Right addr) 1), 1)
1234 pushAtom d p (AnnCast e _)
1235 = pushAtom d p (snd e)
1238 = pprPanic "ByteCodeGen.pushAtom"
1239 (pprCoreExpr (deAnnotate (undefined, other)))
1241 foreign import ccall unsafe "memcpy"
1242 memcpy :: Ptr a -> Ptr b -> CSize -> IO ()
1245 -- -----------------------------------------------------------------------------
1246 -- Given a bunch of alts code and their discrs, do the donkey work
1247 -- of making a multiway branch using a switch tree.
1248 -- What a load of hassle!
1250 mkMultiBranch :: Maybe Int -- # datacons in tycon, if alg alt
1251 -- a hint; generates better code
1252 -- Nothing is always safe
1253 -> [(Discr, BCInstrList)]
1255 mkMultiBranch maybe_ncons raw_ways
1256 = let d_way = filter (isNoDiscr.fst) raw_ways
1258 (\w1 w2 -> leAlt (fst w1) (fst w2))
1259 (filter (not.isNoDiscr.fst) raw_ways)
1261 mkTree :: [(Discr, BCInstrList)] -> Discr -> Discr -> BcM BCInstrList
1262 mkTree [] range_lo range_hi = return the_default
1264 mkTree [val] range_lo range_hi
1265 | range_lo `eqAlt` range_hi
1268 = do label_neq <- getLabelBc
1269 return (mkTestEQ (fst val) label_neq
1271 `appOL` unitOL (LABEL label_neq)
1272 `appOL` the_default))
1274 mkTree vals range_lo range_hi
1275 = let n = length vals `div` 2
1276 vals_lo = take n vals
1277 vals_hi = drop n vals
1278 v_mid = fst (head vals_hi)
1280 label_geq <- getLabelBc
1281 code_lo <- mkTree vals_lo range_lo (dec v_mid)
1282 code_hi <- mkTree vals_hi v_mid range_hi
1283 return (mkTestLT v_mid label_geq
1285 `appOL` unitOL (LABEL label_geq)
1289 = case d_way of [] -> unitOL CASEFAIL
1292 -- None of these will be needed if there are no non-default alts
1293 (mkTestLT, mkTestEQ, init_lo, init_hi)
1295 = panic "mkMultiBranch: awesome foursome"
1297 = case fst (head notd_ways) of {
1298 DiscrI _ -> ( \(DiscrI i) fail_label -> TESTLT_I i fail_label,
1299 \(DiscrI i) fail_label -> TESTEQ_I i fail_label,
1302 DiscrF _ -> ( \(DiscrF f) fail_label -> TESTLT_F f fail_label,
1303 \(DiscrF f) fail_label -> TESTEQ_F f fail_label,
1306 DiscrD _ -> ( \(DiscrD d) fail_label -> TESTLT_D d fail_label,
1307 \(DiscrD d) fail_label -> TESTEQ_D d fail_label,
1310 DiscrP _ -> ( \(DiscrP i) fail_label -> TESTLT_P i fail_label,
1311 \(DiscrP i) fail_label -> TESTEQ_P i fail_label,
1313 DiscrP algMaxBound )
1316 (algMinBound, algMaxBound)
1317 = case maybe_ncons of
1318 Just n -> (0, n - 1)
1319 Nothing -> (minBound, maxBound)
1321 (DiscrI i1) `eqAlt` (DiscrI i2) = i1 == i2
1322 (DiscrF f1) `eqAlt` (DiscrF f2) = f1 == f2
1323 (DiscrD d1) `eqAlt` (DiscrD d2) = d1 == d2
1324 (DiscrP i1) `eqAlt` (DiscrP i2) = i1 == i2
1325 NoDiscr `eqAlt` NoDiscr = True
1328 (DiscrI i1) `leAlt` (DiscrI i2) = i1 <= i2
1329 (DiscrF f1) `leAlt` (DiscrF f2) = f1 <= f2
1330 (DiscrD d1) `leAlt` (DiscrD d2) = d1 <= d2
1331 (DiscrP i1) `leAlt` (DiscrP i2) = i1 <= i2
1332 NoDiscr `leAlt` NoDiscr = True
1335 isNoDiscr NoDiscr = True
1338 dec (DiscrI i) = DiscrI (i-1)
1339 dec (DiscrP i) = DiscrP (i-1)
1340 dec other = other -- not really right, but if you
1341 -- do cases on floating values, you'll get what you deserve
1343 -- same snotty comment applies to the following
1345 minD, maxD :: Double
1351 mkTree notd_ways init_lo init_hi
1354 -- -----------------------------------------------------------------------------
1355 -- Supporting junk for the compilation schemes
1357 -- Describes case alts
1365 instance Outputable Discr where
1366 ppr (DiscrI i) = int i
1367 ppr (DiscrF f) = text (show f)
1368 ppr (DiscrD d) = text (show d)
1369 ppr (DiscrP i) = int i
1370 ppr NoDiscr = text "DEF"
1373 lookupBCEnv_maybe :: BCEnv -> Id -> Maybe Int
1374 lookupBCEnv_maybe = lookupFM
1376 idSizeW :: Id -> Int
1377 idSizeW id = cgRepSizeW (typeCgRep (idType id))
1380 unboxedTupleException :: a
1381 unboxedTupleException
1384 ("Error: bytecode compiler can't handle unboxed tuples.\n"++
1385 " Possibly due to foreign import/export decls in source.\n"++
1386 " Workaround: use -fobject-code, or compile this module to .o separately."))
1389 mkSLIDE n d = if d == 0 then nilOL else unitOL (SLIDE n d)
1392 splitApp :: AnnExpr' id ann -> (AnnExpr' id ann, [AnnExpr' id ann])
1393 -- The arguments are returned in *right-to-left* order
1394 splitApp (AnnApp (_,f) (_,a))
1395 | isTypeAtom a = splitApp f
1396 | otherwise = case splitApp f of
1397 (f', as) -> (f', a:as)
1398 splitApp (AnnNote n (_,e)) = splitApp e
1399 splitApp (AnnCast (_,e) _) = splitApp e
1400 splitApp e = (e, [])
1403 isTypeAtom :: AnnExpr' id ann -> Bool
1404 isTypeAtom (AnnType _) = True
1405 isTypeAtom _ = False
1407 isVoidArgAtom :: AnnExpr' id ann -> Bool
1408 isVoidArgAtom (AnnVar v) = typeCgRep (idType v) == VoidArg
1409 isVoidArgAtom (AnnNote n (_,e)) = isVoidArgAtom e
1410 isVoidArgAtom (AnnCast (_,e) _) = isVoidArgAtom e
1411 isVoidArgAtom _ = False
1413 atomRep :: AnnExpr' Id ann -> CgRep
1414 atomRep (AnnVar v) = typeCgRep (idType v)
1415 atomRep (AnnLit l) = typeCgRep (literalType l)
1416 atomRep (AnnNote n b) = atomRep (snd b)
1417 atomRep (AnnApp f (_, AnnType _)) = atomRep (snd f)
1418 atomRep (AnnLam x e) | isTyVar x = atomRep (snd e)
1419 atomRep (AnnCast b _) = atomRep (snd b)
1420 atomRep other = pprPanic "atomRep" (ppr (deAnnotate (undefined,other)))
1422 isPtrAtom :: AnnExpr' Id ann -> Bool
1423 isPtrAtom e = atomRep e == PtrArg
1425 -- Let szsw be the sizes in words of some items pushed onto the stack,
1426 -- which has initial depth d'. Return the values which the stack environment
1427 -- should map these items to.
1428 mkStackOffsets :: Int -> [Int] -> [Int]
1429 mkStackOffsets original_depth szsw
1430 = map (subtract 1) (tail (scanl (+) original_depth szsw))
1432 -- -----------------------------------------------------------------------------
1433 -- The bytecode generator's monad
1435 type BcPtr = Either ItblPtr (Ptr ())
1439 uniqSupply :: UniqSupply, -- for generating fresh variable names
1440 nextlabel :: Int, -- for generating local labels
1441 malloced :: [BcPtr], -- thunks malloced for current BCO
1442 -- Should be free()d when it is GCd
1443 breakArray :: BreakArray -- array of breakpoint flags
1446 newtype BcM r = BcM (BcM_State -> IO (BcM_State, r))
1448 ioToBc :: IO a -> BcM a
1449 ioToBc io = BcM $ \st -> do
1453 runBc :: UniqSupply -> ModBreaks -> BcM r -> IO (BcM_State, r)
1454 runBc us modBreaks (BcM m)
1455 = m (BcM_State us 0 [] breakArray)
1457 breakArray = modBreaks_flags modBreaks
1459 thenBc :: BcM a -> (a -> BcM b) -> BcM b
1460 thenBc (BcM expr) cont = BcM $ \st0 -> do
1461 (st1, q) <- expr st0
1466 thenBc_ :: BcM a -> BcM b -> BcM b
1467 thenBc_ (BcM expr) (BcM cont) = BcM $ \st0 -> do
1468 (st1, q) <- expr st0
1469 (st2, r) <- cont st1
1472 returnBc :: a -> BcM a
1473 returnBc result = BcM $ \st -> (return (st, result))
1475 instance Monad BcM where
1480 emitBc :: ([BcPtr] -> ProtoBCO Name) -> BcM (ProtoBCO Name)
1482 = BcM $ \st -> return (st{malloced=[]}, bco (malloced st))
1484 recordMallocBc :: Ptr a -> BcM ()
1486 = BcM $ \st -> return (st{malloced = Right (castPtr a) : malloced st}, ())
1488 recordItblMallocBc :: ItblPtr -> BcM ()
1489 recordItblMallocBc a
1490 = BcM $ \st -> return (st{malloced = Left a : malloced st}, ())
1492 getLabelBc :: BcM Int
1494 = BcM $ \st -> return (st{nextlabel = 1 + nextlabel st}, nextlabel st)
1496 getLabelsBc :: Int -> BcM [Int]
1498 = BcM $ \st -> let ctr = nextlabel st
1499 in return (st{nextlabel = ctr+n}, [ctr .. ctr+n-1])
1501 getBreakArray :: BcM BreakArray
1502 getBreakArray = BcM $ \st -> return (st, breakArray st)
1504 newUnique :: BcM Unique
1506 \st -> case splitUniqSupply (uniqSupply st) of
1507 (us1, us2) -> let newState = st { uniqSupply = us2 }
1508 in return (newState, uniqFromSupply us1)
1510 newId :: Type -> BcM Id
1513 return $ mkSysLocal FSLIT("ticked") uniq ty