[Git][ghc/ghc][wip/keepAlive-optionB] 4 commits: Simplify
Ben Gamari
gitlab at gitlab.haskell.org
Thu Sep 10 19:55:56 UTC 2020
Ben Gamari pushed to branch wip/keepAlive-optionB at Glasgow Haskell Compiler / GHC
Commits:
c8a66f71 by GHC GitLab CI at 2020-09-10T19:55:40+00:00
Simplify
- - - - -
8b633c38 by GHC GitLab CI at 2020-09-10T19:55:40+00:00
hadrian: Don't include -fdiagnostics-color in argument hash
Otherwise the input hash will vary with whether colors are requested,
which changed with `isatty`.
Fixes #17983.
- - - - -
8ce97697 by Ben Gamari at 2020-09-10T19:55:40+00:00
base: Use keepAlive# in withForeignPtr
- - - - -
be85911a by GHC GitLab CI at 2020-09-10T19:55:40+00:00
base: Make ForeignPtrContents of ForeignPtr strict
As mentioned in #17290, there are strong benefits to the
`ForeignPtrContents` field of `ForeignPtr` being strict. In particular,
when looking at the reproducer for #17746, I noticed that the
`ForeignPtrContents` was being allocated on every call to
`withForeignPtr` just to be `touch#`'d'. This is a pretty large
overhead for something like `withForeignPtr` that should be free.
This required updating the `bytestring` submodule to avoid a bottoming
`ForeignPtrContents` in `Data.ByteString.Internal`
Fixes #17290.
- - - - -
7 changed files:
- compiler/GHC/Core/Opt/Simplify.hs
- compiler/GHC/CoreToStg/Prep.hs
- hadrian/src/Settings/Builders/Ghc.hs
- hadrian/src/Target.hs
- libraries/base/Foreign/ForeignPtr/Imp.hs
- libraries/base/GHC/ForeignPtr.hs
- libraries/bytestring
Changes:
=====================================
compiler/GHC/Core/Opt/Simplify.hs
=====================================
@@ -68,7 +68,7 @@ import GHC.Utils.Misc
import GHC.Utils.Error
import GHC.Unit.Module ( moduleName, pprModuleName )
import GHC.Core.Multiplicity
-import GHC.Builtin.PrimOps ( PrimOp (SeqOp) )
+import GHC.Builtin.PrimOps ( PrimOp (SeqOp, KeepAliveOp) )
{-
@@ -2014,28 +2014,36 @@ rebuildCall env fun_info
rebuildCall env (ArgInfo { ai_fun = fun, ai_args = rev_args }) cont
= rebuild env (argInfoExpr fun rev_args) cont
+-- | Simplifications of runRW# and keepAlive#
rebuildContOpCall :: SimplEnv -> ArgInfo -> SimplCont -> Maybe (SimplM (SimplFloats, OutExpr))
----------- The runRW# rule. Do this after absorbing all arguments ------
+rebuildContOpCall _env _arg_info cont
+ | contIsStop cont -- Don't fiddle around if the continuation is boring
+ = Nothing
+
-- See Note [Simplification of runRW#] in GHC.CoreToSTG.Prep.
--
--- runRW# :: forall (r :: RuntimeRep) (o :: TYPE r). (State# RealWorld -> o) -> o
--- K[ runRW# rr ty body ] --> runRW rr' ty' (\s. K[ body s ])
+-- N.B. runRW# :: forall (r :: RuntimeRep) (o :: TYPE r).
+-- (State# RealWorld -> o) -> o
+--
+-- K[ runRW# rr ty body ]
+-- ~>
+-- runRW rr' ty' (\s. K[ body s ])
rebuildContOpCall
env
(ArgInfo { ai_fun = fun_id, ai_args = rev_args })
(ApplyToVal { sc_arg = arg, sc_env = arg_se
, sc_cont = cont, sc_hole_ty = fun_ty })
| fun_id `hasKey` runRWKey
- , not (contIsStop cont) -- Don't fiddle around if the continuation is boring
, [ TyArg {}, TyArg {} ] <- rev_args
= Just $
- do { s <- newId (fsLit "s") Many realWorldStatePrimTy
+ do { s <- newId (fsLit "s") One realWorldStatePrimTy
; let (m,_,_) = splitFunTy fun_ty
env' = (arg_se `setInScopeFromE` env) `addNewInScopeIds` [s]
ty' = contResultType cont
+ k'_ty = mkVisFunTy m realWorldStatePrimTy ty'
cont' = ApplyToVal { sc_dup = Simplified, sc_arg = Var s
, sc_env = env', sc_cont = cont
- , sc_hole_ty = mkVisFunTy m realWorldStatePrimTy ty' }
+ , sc_hole_ty = k'_ty }
-- cont' applies to s, then K
; body' <- simplExprC env' arg cont'
; let arg' = Lam s body'
@@ -2043,6 +2051,48 @@ rebuildContOpCall
call' = mkApps (Var fun_id) [mkTyArg rr', mkTyArg ty', arg']
; return (emptyFloats env, call') }
+-- See Note [Simplification of keepAlive#] in GHC.CoreToStg.Prep.
+--
+-- K[keepAlive# @a_rep @a @r_rep @r x s k]
+-- ~>
+-- keepAlive# @a_rep @a @r_rep @r x s K[k]
+rebuildContOpCall
+ env
+ (ArgInfo { ai_fun = fun_id, ai_args = rev_args })
+ (ApplyToVal { sc_arg = k, sc_env = k_se
+ , sc_cont = cont, sc_hole_ty = fun_ty })
+ | Just KeepAliveOp <- isPrimOpId_maybe fun_id
+ , [ ValArg {as_arg=s0}
+ , ValArg {as_arg=x}
+ , TyArg {} -- res_ty
+ , TyArg {} -- res_rep
+ , TyArg {as_arg_ty=arg_ty}
+ , TyArg {as_arg_ty=arg_rep}
+ ] <- rev_args
+ = Just $
+ do { s <- newId (fsLit "s") One realWorldStatePrimTy
+ ; let (m,_,_) = splitFunTy fun_ty
+ k_env = (k_se `setInScopeFromE` env) `addNewInScopeIds` [s]
+ ty' = contResultType cont
+ k'_ty = mkVisFunTy m realWorldStatePrimTy ty'
+ k_cont = ApplyToVal { sc_dup = Simplified, sc_arg = Var s
+ , sc_env = k_env, sc_cont = cont
+ , sc_hole_ty = k'_ty }
+ ; k' <- simplExprC k_env k k_cont
+ ; let env' = zapSubstEnv env
+ ; s0' <- simplExpr env' s0
+ ; x' <- simplExpr env' x
+ ; arg_rep' <- simplType env' arg_rep
+ ; arg_ty' <- simplType env' arg_ty
+ ; let call' = mkApps (Var fun_id)
+ [ mkTyArg arg_rep', mkTyArg arg_ty'
+ , mkTyArg (getRuntimeRep ty'), mkTyArg ty'
+ , x'
+ , s0'
+ , Lam s k'
+ ]
+ ; return (emptyFloats env, call') }
+
rebuildContOpCall _ _ _ = Nothing
{- Note [Trying rewrite rules]
=====================================
compiler/GHC/CoreToStg/Prep.hs
=====================================
@@ -1096,10 +1096,26 @@ Breaking our desired invariant. Ultimately we decided to simply accept that
the continuation may not be a manifest lambda.
+Note [Simplification of keepAlive#]
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+The keepAlive# primop benefits from a similar optimisation to that described in
+Note [Simplification of runRW#] above. Specifically, we transform:
+
+ K[keepAlive# @a_rep @a @r_rep @r x s k]
+ ~>
+ keepAlive# @a_rep @a @r_rep @r x s K[k]
+
+The reasons are similar to those described in Note [Simplification of runRW#].
+
+-}
+
+
-- ---------------------------------------------------------------------------
-- CpeArg: produces a result satisfying CpeArg
-- ---------------------------------------------------------------------------
+{-
Note [ANF-ising literal string arguments]
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
=====================================
hadrian/src/Settings/Builders/Ghc.hs
=====================================
@@ -35,6 +35,9 @@ compileAndLinkHs = (builder (Ghc CompileHs) ||^ builder (Ghc LinkHs)) ? do
hasDynamic = elem dynamic ways
mconcat [ arg "-Wall"
, not useColor ? builder (Ghc CompileHs) ?
+ -- N.B. Target.trackArgument ignores this argument from the
+ -- input hash to avoid superfluous recompilation, avoiding
+ -- #18672.
arg "-fdiagnostics-color=never"
, (hasVanilla && hasDynamic) ? builder (Ghc CompileHs) ?
platformSupportsSharedLibs ? way vanilla ?
=====================================
hadrian/src/Target.hs
=====================================
@@ -21,11 +21,12 @@ type Target = H.Target Context Builder
trackArgument :: Target -> String -> Bool
trackArgument target arg = case builder target of
Make _ -> not $ threadArg arg
- Ghc _ _ -> not $ verbosityArg arg
+ Ghc _ _ -> not $ verbosityArg arg || diagnosticsColorArg arg
Cabal _ _ -> not $ verbosityArg arg || cabal_configure_ignore arg
_ -> True
where
threadArg s = dropWhileEnd isDigit s `elem` ["-j", "MAKEFLAGS=-j", "THREADS="]
verbosityArg s = dropWhileEnd isDigit s == "-v"
+ diagnosticsColorArg s = "-fdiagnostics-color=" `isPrefixOf` s -- N.B. #18672
cabal_configure_ignore s =
s `elem` [ "--configure-option=--quiet", "--configure-option=--disable-option-checking" ]
=====================================
libraries/base/Foreign/ForeignPtr/Imp.hs
=====================================
@@ -66,31 +66,6 @@ newForeignPtr finalizer p
addForeignPtrFinalizer finalizer fObj
return fObj
-withForeignPtr :: ForeignPtr a -> (Ptr a -> IO b) -> IO b
--- ^This is a way to look at the pointer living inside a
--- foreign object. This function takes a function which is
--- applied to that pointer. The resulting 'IO' action is then
--- executed. The foreign object is kept alive at least during
--- the whole action, even if it is not used directly
--- inside. Note that it is not safe to return the pointer from
--- the action and use it after the action completes. All uses
--- of the pointer should be inside the
--- 'withForeignPtr' bracket. The reason for
--- this unsafeness is the same as for
--- 'unsafeForeignPtrToPtr' below: the finalizer
--- may run earlier than expected, because the compiler can only
--- track usage of the 'ForeignPtr' object, not
--- a 'Ptr' object made from it.
---
--- This function is normally used for marshalling data to
--- or from the object pointed to by the
--- 'ForeignPtr', using the operations from the
--- 'Storable' class.
-withForeignPtr fo io
- = do r <- io (unsafeForeignPtrToPtr fo)
- touchForeignPtr fo
- return r
-
-- | This variant of 'newForeignPtr' adds a finalizer that expects an
-- environment in addition to the finalized pointer. The environment
-- that will be passed to the finalizer is fixed by the second argument to
=====================================
libraries/base/GHC/ForeignPtr.hs
=====================================
@@ -46,6 +46,7 @@ module GHC.ForeignPtr
castForeignPtr,
plusForeignPtr,
-- * Finalization
+ withForeignPtr,
touchForeignPtr,
finalizeForeignPtr
-- * Commentary
@@ -79,7 +80,7 @@ import Unsafe.Coerce ( unsafeCoerce, unsafeCoerceUnlifted )
-- type argument of 'ForeignPtr' should normally be an instance of
-- class 'Storable'.
--
-data ForeignPtr a = ForeignPtr Addr# ForeignPtrContents
+data ForeignPtr a = ForeignPtr Addr# !ForeignPtrContents
-- The Addr# in the ForeignPtr object is intentionally stored
-- separately from the finalizer. The primary aim of the
-- representation is to make withForeignPtr efficient; in fact,
@@ -503,6 +504,31 @@ newForeignPtr_ (Ptr obj) = do
r <- newIORef NoFinalizers
return (ForeignPtr obj (PlainForeignPtr r))
+withForeignPtr :: ForeignPtr a -> (Ptr a -> IO b) -> IO b
+-- ^This is a way to look at the pointer living inside a
+-- foreign object. This function takes a function which is
+-- applied to that pointer. The resulting 'IO' action is then
+-- executed. The foreign object is kept alive at least during
+-- the whole action, even if it is not used directly
+-- inside. Note that it is not safe to return the pointer from
+-- the action and use it after the action completes. All uses
+-- of the pointer should be inside the
+-- 'withForeignPtr' bracket. The reason for
+-- this unsafeness is the same as for
+-- 'unsafeForeignPtrToPtr' below: the finalizer
+-- may run earlier than expected, because the compiler can only
+-- track usage of the 'ForeignPtr' object, not
+-- a 'Ptr' object made from it.
+--
+-- This function is normally used for marshalling data to
+-- or from the object pointed to by the
+-- 'ForeignPtr', using the operations from the
+-- 'Storable' class.
+withForeignPtr fo@(ForeignPtr _ r) f = IO $ \s ->
+ case f (unsafeForeignPtrToPtr fo) of
+ IO action# -> keepAlive# r s action#
+
+
touchForeignPtr :: ForeignPtr a -> IO ()
-- ^This function ensures that the foreign object in
-- question is alive at the given place in the sequence of IO
=====================================
libraries/bytestring
=====================================
@@ -1 +1 @@
-Subproject commit e6cb01e2ec0bfdd19298418c85f220925a9fa307
+Subproject commit cff1ad3422ed463cf92ea996b276da9156200bbf
View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/compare/36c95e394bd58147070966409d4ba926799db094...be85911a5946ad341e1cca1684655619bec15339
--
View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/compare/36c95e394bd58147070966409d4ba926799db094...be85911a5946ad341e1cca1684655619bec15339
You're receiving this email because of your account on gitlab.haskell.org.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.haskell.org/pipermail/ghc-commits/attachments/20200910/00419e32/attachment-0001.html>
More information about the ghc-commits
mailing list