Repository : ssh://darcs.haskell.org//srv/darcs/ghc On branch : master
http://hackage.haskell.org/trac/ghc/changeset/6181e007f0e1e8eddba7acf0d5fbcbaf46806249 >--------------------------------------------------------------- commit 6181e007f0e1e8eddba7acf0d5fbcbaf46806249 Author: Simon Marlow <[email protected]> Date: Fri Jun 15 10:30:35 2012 +0100 Don't put auto sccs on INLINE functions (#6131) There was also a bug caused by INLINEs getting SCCs, but unfortunately I have lost the test case. The Note in the code describes the problem though. >--------------------------------------------------------------- compiler/deSugar/Coverage.lhs | 36 +++++++++++++++++++++++++++++++++--- 1 files changed, 33 insertions(+), 3 deletions(-) diff --git a/compiler/deSugar/Coverage.lhs b/compiler/deSugar/Coverage.lhs index fa7c343..c29f39e 100644 --- a/compiler/deSugar/Coverage.lhs +++ b/compiler/deSugar/Coverage.lhs @@ -84,6 +84,7 @@ addTicksToBinds dflags mod mod_loc exports tyCons binds = , declPath = [] , tte_dflags = dflags , exports = exports + , inlines = emptyVarSet , inScope = emptyVarSet , blackList = Map.fromList [ (getSrcSpan (tyConName tyCon),()) @@ -231,6 +232,7 @@ addTickLHsBind :: LHsBind Id -> TM (LHsBind Id) addTickLHsBind (L pos bind@(AbsBinds { abs_binds = binds, abs_exports = abs_exports })) = do withEnv add_exports $ do + withEnv add_inlines $ do binds' <- addTickLHsBinds binds return $ L pos $ bind { abs_binds = binds' } where @@ -245,9 +247,24 @@ addTickLHsBind (L pos bind@(AbsBinds { abs_binds = binds, | ABE{ abe_poly = pid, abe_mono = mid } <- abs_exports , idName pid `elemNameSet` (exports env) ] } + add_inlines env = + env{ inlines = inlines env `extendVarSetList` + [ mid + | ABE{ abe_poly = pid, abe_mono = mid } <- abs_exports + , isAnyInlinePragma (idInlinePragma pid) ] } + + addTickLHsBind (L pos (funBind@(FunBind { fun_id = (L _ id) }))) = do let name = getOccString id decl_path <- getPathEntry + density <- getDensity + + inline_ids <- liftM inlines getEnv + let inline = isAnyInlinePragma (idInlinePragma id) + || id `elemVarSet` inline_ids + + -- See Note [inline sccs] + if inline && opt_SccProfilingOn then return (L pos funBind) else do (fvs, (MatchGroup matches' ty)) <- getFreeVars $ @@ -255,7 +272,6 @@ addTickLHsBind (L pos (funBind@(FunBind { fun_id = (L _ id) }))) = do addTickMatchGroup False (fun_matches funBind) blackListed <- isBlackListed pos - density <- getDensity exported_names <- liftM exports getEnv -- We don't want to generate code for blacklisted positions @@ -264,8 +280,6 @@ addTickLHsBind (L pos (funBind@(FunBind { fun_id = (L _ id) }))) = do let simple = isSimplePatBind funBind toplev = null decl_path exported = idName id `elemNameSet` exported_names - inline = {- pprTrace "inline" (ppr id <+> ppr (idInlinePragma id)) $ -} - isAnyInlinePragma (idInlinePragma id) tick <- if not blackListed && shouldTickBind density toplev exported simple inline @@ -321,6 +335,21 @@ bindTick density name pos fvs = do allocATickBox box_label count_entries top_only pos fvs +-- Note [inline sccs] +-- +-- It should be reasonable to add ticks to INLINE functions; however +-- currently this tickles a bug later on because the SCCfinal pass +-- does not look inside unfoldings to find CostCentres. It would be +-- difficult to fix that, because SCCfinal currently works on STG and +-- not Core (and since it also generates CostCentres for CAFs, +-- changing this would be difficult too). +-- +-- Another reason not to add ticks to INLINE functions is that this +-- sometimes handy for avoiding adding a tick to a particular function +-- (see #6131) +-- +-- So for now we do not add any ticks to INLINE functions at all. + -- ----------------------------------------------------------------------------- -- Decorate an LHsExpr with ticks @@ -869,6 +898,7 @@ data TickTransEnv = TTE { fileName :: FastString , density :: TickDensity , tte_dflags :: DynFlags , exports :: NameSet + , inlines :: VarSet , declPath :: [String] , inScope :: VarSet , blackList :: Map SrcSpan () _______________________________________________ Cvs-ghc mailing list [email protected] http://www.haskell.org/mailman/listinfo/cvs-ghc
