{-# LANGUAGE BangPatterns #-}
{-# LANGUAGE LambdaCase #-}
{-# LANGUAGE TupleSections #-}

module GHC.Cmm.Pipeline (
  -- | Converts C-- with an implicit stack and native C-- calls into
  -- optimized, CPS converted and native-call-less C--.  The latter
  -- C-- can be used to generate assembly.
  cmmPipeline
) where

import GHC.Prelude

import GHC.Cmm
import GHC.Cmm.Lint
import GHC.Cmm.Info.Build
import GHC.Cmm.CommonBlockElim
import GHC.Cmm.Switch.Implement
import GHC.Cmm.ProcPoint
import GHC.Cmm.ContFlowOpt
import GHC.Cmm.LayoutStack
import GHC.Cmm.Sink
import GHC.Cmm.Dataflow.Collections

import GHC.Types.Unique.Supply
import GHC.Driver.Session
import GHC.Utils.Error
import GHC.Driver.Types
import Control.Monad
import GHC.Utils.Outputable
import GHC.Platform
import Data.Either (partitionEithers)

-----------------------------------------------------------------------------
-- | Top level driver for C-- pipeline
-----------------------------------------------------------------------------

cmmPipeline
 :: HscEnv -- Compilation env including
           -- dynamic flags: -dcmm-lint -ddump-cmm-cps
 -> ModuleSRTInfo        -- Info about SRTs generated so far
 -> CmmGroup             -- Input C-- with Procedures
 -> IO (ModuleSRTInfo, CmmGroupSRTs) -- Output CPS transformed C--

cmmPipeline :: HscEnv
-> ModuleSRTInfo -> CmmGroup -> IO (ModuleSRTInfo, [CmmDeclSRTs])
cmmPipeline HscEnv
hsc_env ModuleSRTInfo
srtInfo CmmGroup
prog = DynFlags
-> SDoc
-> ((ModuleSRTInfo, [CmmDeclSRTs]) -> ())
-> IO (ModuleSRTInfo, [CmmDeclSRTs])
-> IO (ModuleSRTInfo, [CmmDeclSRTs])
forall (m :: * -> *) a.
MonadIO m =>
DynFlags -> SDoc -> (a -> ()) -> m a -> m a
withTimingSilent DynFlags
dflags (String -> SDoc
text String
"Cmm pipeline") (ModuleSRTInfo, [CmmDeclSRTs]) -> ()
forall {t :: * -> *} {a} {a}. Foldable t => (a, t a) -> ()
forceRes (IO (ModuleSRTInfo, [CmmDeclSRTs])
 -> IO (ModuleSRTInfo, [CmmDeclSRTs]))
-> IO (ModuleSRTInfo, [CmmDeclSRTs])
-> IO (ModuleSRTInfo, [CmmDeclSRTs])
forall a b. (a -> b) -> a -> b
$
  do let dflags :: DynFlags
dflags = HscEnv -> DynFlags
hsc_dflags HscEnv
hsc_env

     [Either (CAFEnv, CmmGroup) (Set CAFLabel, CmmDecl)]
tops <- {-# SCC "tops" #-} (CmmDecl -> IO (Either (CAFEnv, CmmGroup) (Set CAFLabel, CmmDecl)))
-> CmmGroup
-> IO [Either (CAFEnv, CmmGroup) (Set CAFLabel, CmmDecl)]
forall (t :: * -> *) (m :: * -> *) a b.
(Traversable t, Monad m) =>
(a -> m b) -> t a -> m (t b)
mapM (HscEnv
-> CmmDecl
-> IO (Either (CAFEnv, CmmGroup) (Set CAFLabel, CmmDecl))
cpsTop HscEnv
hsc_env) CmmGroup
prog

     let ([(CAFEnv, CmmGroup)]
procs, [(Set CAFLabel, CmmDecl)]
data_) = [Either (CAFEnv, CmmGroup) (Set CAFLabel, CmmDecl)]
-> ([(CAFEnv, CmmGroup)], [(Set CAFLabel, CmmDecl)])
forall a b. [Either a b] -> ([a], [b])
partitionEithers [Either (CAFEnv, CmmGroup) (Set CAFLabel, CmmDecl)]
tops
     (ModuleSRTInfo
srtInfo, [CmmDeclSRTs]
cmms) <- {-# SCC "doSRTs" #-} DynFlags
-> ModuleSRTInfo
-> [(CAFEnv, CmmGroup)]
-> [(Set CAFLabel, CmmDecl)]
-> IO (ModuleSRTInfo, [CmmDeclSRTs])
doSRTs DynFlags
dflags ModuleSRTInfo
srtInfo [(CAFEnv, CmmGroup)]
procs [(Set CAFLabel, CmmDecl)]
data_
     DynFlags -> DumpFlag -> String -> DumpFormat -> SDoc -> IO ()
dumpWith DynFlags
dflags DumpFlag
Opt_D_dump_cmm_cps String
"Post CPS Cmm" DumpFormat
FormatCMM ([CmmDeclSRTs] -> SDoc
forall a. Outputable a => a -> SDoc
ppr [CmmDeclSRTs]
cmms)

     (ModuleSRTInfo, [CmmDeclSRTs]) -> IO (ModuleSRTInfo, [CmmDeclSRTs])
forall (m :: * -> *) a. Monad m => a -> m a
return (ModuleSRTInfo
srtInfo, [CmmDeclSRTs]
cmms)

  where forceRes :: (a, t a) -> ()
forceRes (a
info, t a
group) =
          a
info a -> () -> ()
`seq` (a -> () -> ()) -> () -> t a -> ()
forall (t :: * -> *) a b.
Foldable t =>
(a -> b -> b) -> b -> t a -> b
foldr (\a
decl ()
r -> a
decl a -> () -> ()
`seq` ()
r) () t a
group

        dflags :: DynFlags
dflags = HscEnv -> DynFlags
hsc_dflags HscEnv
hsc_env

cpsTop :: HscEnv -> CmmDecl -> IO (Either (CAFEnv, [CmmDecl]) (CAFSet, CmmDecl))
cpsTop :: HscEnv
-> CmmDecl
-> IO (Either (CAFEnv, CmmGroup) (Set CAFLabel, CmmDecl))
cpsTop HscEnv
_ p :: CmmDecl
p@(CmmData Section
_ GenCmmStatics 'False
statics) = Either (CAFEnv, CmmGroup) (Set CAFLabel, CmmDecl)
-> IO (Either (CAFEnv, CmmGroup) (Set CAFLabel, CmmDecl))
forall (m :: * -> *) a. Monad m => a -> m a
return ((Set CAFLabel, CmmDecl)
-> Either (CAFEnv, CmmGroup) (Set CAFLabel, CmmDecl)
forall a b. b -> Either a b
Right (GenCmmStatics 'False -> Set CAFLabel
cafAnalData GenCmmStatics 'False
statics, CmmDecl
p))
cpsTop HscEnv
hsc_env CmmDecl
proc =
    do
       ----------- Control-flow optimisations ----------------------------------

       -- The first round of control-flow optimisation speeds up the
       -- later passes by removing lots of empty blocks, so we do it
       -- even when optimisation isn't turned on.
       --
       CmmProc CmmTopInfo
h CLabel
l [GlobalReg]
v CmmGraph
g <- {-# SCC "cmmCfgOpts(1)" #-}
            CmmDecl -> IO CmmDecl
forall (m :: * -> *) a. Monad m => a -> m a
return (CmmDecl -> IO CmmDecl) -> CmmDecl -> IO CmmDecl
forall a b. (a -> b) -> a -> b
$ Bool -> CmmDecl -> CmmDecl
cmmCfgOptsProc Bool
splitting_proc_points CmmDecl
proc
       DumpFlag -> String -> CmmGraph -> IO ()
dump DumpFlag
Opt_D_dump_cmm_cfg String
"Post control-flow optimisations" CmmGraph
g

       let !TopInfo {stack_info :: CmmTopInfo -> CmmStackInfo
stack_info=StackInfo { arg_space :: CmmStackInfo -> Int
arg_space = Int
entry_off
                                          , do_layout :: CmmStackInfo -> Bool
do_layout = Bool
do_layout }} = CmmTopInfo
h

       ----------- Eliminate common blocks -------------------------------------
       CmmGraph
g <- {-# SCC "elimCommonBlocks" #-}
            GeneralFlag
-> (CmmGraph -> CmmGraph)
-> CmmGraph
-> DumpFlag
-> String
-> IO CmmGraph
condPass GeneralFlag
Opt_CmmElimCommonBlocks CmmGraph -> CmmGraph
elimCommonBlocks CmmGraph
g
                          DumpFlag
Opt_D_dump_cmm_cbe String
"Post common block elimination"

       -- Any work storing block Labels must be performed _after_
       -- elimCommonBlocks

       ----------- Implement switches ------------------------------------------
       CmmGraph
g <- {-# SCC "createSwitchPlans" #-}
            UniqSM CmmGraph -> IO CmmGraph
forall a. UniqSM a -> IO a
runUniqSM (UniqSM CmmGraph -> IO CmmGraph) -> UniqSM CmmGraph -> IO CmmGraph
forall a b. (a -> b) -> a -> b
$ DynFlags -> CmmGraph -> UniqSM CmmGraph
cmmImplementSwitchPlans DynFlags
dflags CmmGraph
g
       DumpFlag -> String -> CmmGraph -> IO ()
dump DumpFlag
Opt_D_dump_cmm_switch String
"Post switch plan" CmmGraph
g

       ----------- Proc points -------------------------------------------------
       let
         call_pps :: ProcPointSet -- LabelMap
         call_pps :: ProcPointSet
call_pps = {-# SCC "callProcPoints" #-} CmmGraph -> ProcPointSet
callProcPoints CmmGraph
g
       ProcPointSet
proc_points <-
          if Bool
splitting_proc_points
             then do
               ProcPointSet
pp <- {-# SCC "minimalProcPointSet" #-} UniqSM ProcPointSet -> IO ProcPointSet
forall a. UniqSM a -> IO a
runUniqSM (UniqSM ProcPointSet -> IO ProcPointSet)
-> UniqSM ProcPointSet -> IO ProcPointSet
forall a b. (a -> b) -> a -> b
$
                  Platform -> ProcPointSet -> CmmGraph -> UniqSM ProcPointSet
minimalProcPointSet (DynFlags -> Platform
targetPlatform DynFlags
dflags) ProcPointSet
call_pps CmmGraph
g
               DynFlags -> DumpFlag -> String -> DumpFormat -> SDoc -> IO ()
dumpWith DynFlags
dflags DumpFlag
Opt_D_dump_cmm_proc String
"Proc points"
                     DumpFormat
FormatCMM (CLabel -> SDoc
forall a. Outputable a => a -> SDoc
ppr CLabel
l SDoc -> SDoc -> SDoc
$$ ProcPointSet -> SDoc
forall a. Outputable a => a -> SDoc
ppr ProcPointSet
pp SDoc -> SDoc -> SDoc
$$ CmmGraph -> SDoc
forall a. Outputable a => a -> SDoc
ppr CmmGraph
g)
               ProcPointSet -> IO ProcPointSet
forall (m :: * -> *) a. Monad m => a -> m a
return ProcPointSet
pp
             else
               ProcPointSet -> IO ProcPointSet
forall (m :: * -> *) a. Monad m => a -> m a
return ProcPointSet
call_pps

       ----------- Layout the stack and manifest Sp ----------------------------
       (CmmGraph
g, LabelMap StackMap
stackmaps) <-
            {-# SCC "layoutStack" #-}
            if Bool
do_layout
               then UniqSM (CmmGraph, LabelMap StackMap)
-> IO (CmmGraph, LabelMap StackMap)
forall a. UniqSM a -> IO a
runUniqSM (UniqSM (CmmGraph, LabelMap StackMap)
 -> IO (CmmGraph, LabelMap StackMap))
-> UniqSM (CmmGraph, LabelMap StackMap)
-> IO (CmmGraph, LabelMap StackMap)
forall a b. (a -> b) -> a -> b
$ DynFlags
-> ProcPointSet
-> Int
-> CmmGraph
-> UniqSM (CmmGraph, LabelMap StackMap)
cmmLayoutStack DynFlags
dflags ProcPointSet
proc_points Int
entry_off CmmGraph
g
               else (CmmGraph, LabelMap StackMap) -> IO (CmmGraph, LabelMap StackMap)
forall (m :: * -> *) a. Monad m => a -> m a
return (CmmGraph
g, LabelMap StackMap
forall (map :: * -> *) a. IsMap map => map a
mapEmpty)
       DumpFlag -> String -> CmmGraph -> IO ()
dump DumpFlag
Opt_D_dump_cmm_sp String
"Layout Stack" CmmGraph
g

       ----------- Sink and inline assignments  --------------------------------
       CmmGraph
g <- {-# SCC "sink" #-} -- See Note [Sinking after stack layout]
            GeneralFlag
-> (CmmGraph -> CmmGraph)
-> CmmGraph
-> DumpFlag
-> String
-> IO CmmGraph
condPass GeneralFlag
Opt_CmmSink (DynFlags -> CmmGraph -> CmmGraph
cmmSink DynFlags
dflags) CmmGraph
g
                     DumpFlag
Opt_D_dump_cmm_sink String
"Sink assignments"

       ------------- CAF analysis ----------------------------------------------
       let cafEnv :: CAFEnv
cafEnv = {-# SCC "cafAnal" #-} ProcPointSet -> CLabel -> CmmGraph -> CAFEnv
cafAnal ProcPointSet
call_pps CLabel
l CmmGraph
g
       DynFlags -> DumpFlag -> String -> DumpFormat -> SDoc -> IO ()
dumpWith DynFlags
dflags DumpFlag
Opt_D_dump_cmm_caf String
"CAFEnv" DumpFormat
FormatText (CAFEnv -> SDoc
forall a. Outputable a => a -> SDoc
ppr CAFEnv
cafEnv)

       CmmGroup
g <- if Bool
splitting_proc_points
            then do
               ------------- Split into separate procedures -----------------------
               let pp_map :: LabelMap Status
pp_map = {-# SCC "procPointAnalysis" #-}
                            ProcPointSet -> CmmGraph -> LabelMap Status
procPointAnalysis ProcPointSet
proc_points CmmGraph
g
               DynFlags -> DumpFlag -> String -> DumpFormat -> SDoc -> IO ()
dumpWith DynFlags
dflags DumpFlag
Opt_D_dump_cmm_procmap String
"procpoint map"
                  DumpFormat
FormatCMM (LabelMap Status -> SDoc
forall a. Outputable a => a -> SDoc
ppr LabelMap Status
pp_map)
               CmmGroup
g <- {-# SCC "splitAtProcPoints" #-} UniqSM CmmGroup -> IO CmmGroup
forall a. UniqSM a -> IO a
runUniqSM (UniqSM CmmGroup -> IO CmmGroup) -> UniqSM CmmGroup -> IO CmmGroup
forall a b. (a -> b) -> a -> b
$
                    DynFlags
-> CLabel
-> ProcPointSet
-> ProcPointSet
-> LabelMap Status
-> CmmDecl
-> UniqSM CmmGroup
splitAtProcPoints DynFlags
dflags CLabel
l ProcPointSet
call_pps ProcPointSet
proc_points LabelMap Status
pp_map
                                      (CmmTopInfo -> CLabel -> [GlobalReg] -> CmmGraph -> CmmDecl
forall d h g. h -> CLabel -> [GlobalReg] -> g -> GenCmmDecl d h g
CmmProc CmmTopInfo
h CLabel
l [GlobalReg]
v CmmGraph
g)
               DumpFlag -> String -> CmmGroup -> IO ()
forall {t :: * -> *} {a}.
(Foldable t, Outputable a) =>
DumpFlag -> String -> t a -> IO ()
dumps DumpFlag
Opt_D_dump_cmm_split String
"Post splitting" CmmGroup
g
               CmmGroup -> IO CmmGroup
forall (m :: * -> *) a. Monad m => a -> m a
return CmmGroup
g
             else do
               -- attach info tables to return points
               CmmGroup -> IO CmmGroup
forall (m :: * -> *) a. Monad m => a -> m a
return (CmmGroup -> IO CmmGroup) -> CmmGroup -> IO CmmGroup
forall a b. (a -> b) -> a -> b
$ [ProcPointSet -> CmmDecl -> CmmDecl
attachContInfoTables ProcPointSet
call_pps (CmmTopInfo -> CLabel -> [GlobalReg] -> CmmGraph -> CmmDecl
forall d h g. h -> CLabel -> [GlobalReg] -> g -> GenCmmDecl d h g
CmmProc CmmTopInfo
h CLabel
l [GlobalReg]
v CmmGraph
g)]

       ------------- Populate info tables with stack info -----------------
       CmmGroup
g <- {-# SCC "setInfoTableStackMap" #-}
            CmmGroup -> IO CmmGroup
forall (m :: * -> *) a. Monad m => a -> m a
return (CmmGroup -> IO CmmGroup) -> CmmGroup -> IO CmmGroup
forall a b. (a -> b) -> a -> b
$ (CmmDecl -> CmmDecl) -> CmmGroup -> CmmGroup
forall a b. (a -> b) -> [a] -> [b]
map (Platform -> LabelMap StackMap -> CmmDecl -> CmmDecl
setInfoTableStackMap Platform
platform LabelMap StackMap
stackmaps) CmmGroup
g
       DumpFlag -> String -> CmmGroup -> IO ()
forall {t :: * -> *} {a}.
(Foldable t, Outputable a) =>
DumpFlag -> String -> t a -> IO ()
dumps DumpFlag
Opt_D_dump_cmm_info String
"after setInfoTableStackMap" CmmGroup
g

       ----------- Control-flow optimisations -----------------------------
       CmmGroup
g <- {-# SCC "cmmCfgOpts(2)" #-}
            CmmGroup -> IO CmmGroup
forall (m :: * -> *) a. Monad m => a -> m a
return (CmmGroup -> IO CmmGroup) -> CmmGroup -> IO CmmGroup
forall a b. (a -> b) -> a -> b
$ if DynFlags -> Int
optLevel DynFlags
dflags Int -> Int -> Bool
forall a. Ord a => a -> a -> Bool
>= Int
1
                     then (CmmDecl -> CmmDecl) -> CmmGroup -> CmmGroup
forall a b. (a -> b) -> [a] -> [b]
map (Bool -> CmmDecl -> CmmDecl
cmmCfgOptsProc Bool
splitting_proc_points) CmmGroup
g
                     else CmmGroup
g
       CmmGroup
g <- CmmGroup -> IO CmmGroup
forall (m :: * -> *) a. Monad m => a -> m a
return ((CmmDecl -> CmmDecl) -> CmmGroup -> CmmGroup
forall a b. (a -> b) -> [a] -> [b]
map CmmDecl -> CmmDecl
removeUnreachableBlocksProc CmmGroup
g)
            -- See Note [unreachable blocks]
       DumpFlag -> String -> CmmGroup -> IO ()
forall {t :: * -> *} {a}.
(Foldable t, Outputable a) =>
DumpFlag -> String -> t a -> IO ()
dumps DumpFlag
Opt_D_dump_cmm_cfg String
"Post control-flow optimisations" CmmGroup
g

       Either (CAFEnv, CmmGroup) (Set CAFLabel, CmmDecl)
-> IO (Either (CAFEnv, CmmGroup) (Set CAFLabel, CmmDecl))
forall (m :: * -> *) a. Monad m => a -> m a
return ((CAFEnv, CmmGroup)
-> Either (CAFEnv, CmmGroup) (Set CAFLabel, CmmDecl)
forall a b. a -> Either a b
Left (CAFEnv
cafEnv, CmmGroup
g))

  where dflags :: DynFlags
dflags = HscEnv -> DynFlags
hsc_dflags HscEnv
hsc_env
        platform :: Platform
platform = DynFlags -> Platform
targetPlatform DynFlags
dflags
        dump :: DumpFlag -> String -> CmmGraph -> IO ()
dump = DynFlags -> DumpFlag -> String -> CmmGraph -> IO ()
dumpGraph DynFlags
dflags

        dumps :: DumpFlag -> String -> t a -> IO ()
dumps DumpFlag
flag String
name
           = (a -> IO ()) -> t a -> IO ()
forall (t :: * -> *) (m :: * -> *) a b.
(Foldable t, Monad m) =>
(a -> m b) -> t a -> m ()
mapM_ (DynFlags -> DumpFlag -> String -> DumpFormat -> SDoc -> IO ()
dumpWith DynFlags
dflags DumpFlag
flag String
name DumpFormat
FormatCMM (SDoc -> IO ()) -> (a -> SDoc) -> a -> IO ()
forall b c a. (b -> c) -> (a -> b) -> a -> c
. a -> SDoc
forall a. Outputable a => a -> SDoc
ppr)

        condPass :: GeneralFlag
-> (CmmGraph -> CmmGraph)
-> CmmGraph
-> DumpFlag
-> String
-> IO CmmGraph
condPass GeneralFlag
flag CmmGraph -> CmmGraph
pass CmmGraph
g DumpFlag
dumpflag String
dumpname =
            if GeneralFlag -> DynFlags -> Bool
gopt GeneralFlag
flag DynFlags
dflags
               then do
                    CmmGraph
g <- CmmGraph -> IO CmmGraph
forall (m :: * -> *) a. Monad m => a -> m a
return (CmmGraph -> IO CmmGraph) -> CmmGraph -> IO CmmGraph
forall a b. (a -> b) -> a -> b
$ CmmGraph -> CmmGraph
pass CmmGraph
g
                    DumpFlag -> String -> CmmGraph -> IO ()
dump DumpFlag
dumpflag String
dumpname CmmGraph
g
                    CmmGraph -> IO CmmGraph
forall (m :: * -> *) a. Monad m => a -> m a
return CmmGraph
g
               else CmmGraph -> IO CmmGraph
forall (m :: * -> *) a. Monad m => a -> m a
return CmmGraph
g

        -- we don't need to split proc points for the NCG, unless
        -- tablesNextToCode is off.  The latter is because we have no
        -- label to put on info tables for basic blocks that are not
        -- the entry point.
        splitting_proc_points :: Bool
splitting_proc_points = DynFlags -> HscTarget
hscTarget DynFlags
dflags HscTarget -> HscTarget -> Bool
forall a. Eq a => a -> a -> Bool
/= HscTarget
HscAsm
                             Bool -> Bool -> Bool
|| Bool -> Bool
not (Platform -> Bool
platformTablesNextToCode Platform
platform)
                             Bool -> Bool -> Bool
|| -- Note [inconsistent-pic-reg]
                                Bool
usingInconsistentPicReg
        usingInconsistentPicReg :: Bool
usingInconsistentPicReg
           = case (Platform -> Arch
platformArch Platform
platform, Platform -> OS
platformOS Platform
platform, DynFlags -> Bool
positionIndependent DynFlags
dflags)
             of   (Arch
ArchX86, OS
OSDarwin, Bool
pic) -> Bool
pic
                  (Arch, OS, Bool)
_                        -> Bool
False

-- Note [Sinking after stack layout]
-- ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
--
-- In the past we considered running sinking pass also before stack
-- layout, but after making some measurements we realized that:
--
--   a) running sinking only before stack layout produces slower
--      code than running sinking only before stack layout
--
--   b) running sinking both before and after stack layout produces
--      code that has the same performance as when running sinking
--      only after stack layout.
--
-- In other words sinking before stack layout doesn't buy as anything.
--
-- An interesting question is "why is it better to run sinking after
-- stack layout"? It seems that the major reason are stores and loads
-- generated by stack layout. Consider this code before stack layout:
--
--  c1E:
--      _c1C::P64 = R3;
--      _c1B::P64 = R2;
--      _c1A::P64 = R1;
--      I64[(young<c1D> + 8)] = c1D;
--      call stg_gc_noregs() returns to c1D, args: 8, res: 8, upd: 8;
--  c1D:
--      R3 = _c1C::P64;
--      R2 = _c1B::P64;
--      R1 = _c1A::P64;
--      call (P64[(old + 8)])(R3, R2, R1) args: 8, res: 0, upd: 8;
--
-- Stack layout pass will save all local variables live across a call
-- (_c1C, _c1B and _c1A in this example) on the stack just before
-- making a call and reload them from the stack after returning from a
-- call:
--
--  c1E:
--      _c1C::P64 = R3;
--      _c1B::P64 = R2;
--      _c1A::P64 = R1;
--      I64[Sp - 32] = c1D;
--      P64[Sp - 24] = _c1A::P64;
--      P64[Sp - 16] = _c1B::P64;
--      P64[Sp - 8] = _c1C::P64;
--      Sp = Sp - 32;
--      call stg_gc_noregs() returns to c1D, args: 8, res: 8, upd: 8;
--  c1D:
--      _c1A::P64 = P64[Sp + 8];
--      _c1B::P64 = P64[Sp + 16];
--      _c1C::P64 = P64[Sp + 24];
--      R3 = _c1C::P64;
--      R2 = _c1B::P64;
--      R1 = _c1A::P64;
--      Sp = Sp + 32;
--      call (P64[Sp])(R3, R2, R1) args: 8, res: 0, upd: 8;
--
-- If we don't run sinking pass after stack layout we are basically
-- left with such code. However, running sinking on this code can lead
-- to significant improvements:
--
--  c1E:
--      I64[Sp - 32] = c1D;
--      P64[Sp - 24] = R1;
--      P64[Sp - 16] = R2;
--      P64[Sp - 8] = R3;
--      Sp = Sp - 32;
--      call stg_gc_noregs() returns to c1D, args: 8, res: 8, upd: 8;
--  c1D:
--      R3 = P64[Sp + 24];
--      R2 = P64[Sp + 16];
--      R1 = P64[Sp + 8];
--      Sp = Sp + 32;
--      call (P64[Sp])(R3, R2, R1) args: 8, res: 0, upd: 8;
--
-- Now we only have 9 assignments instead of 15.
--
-- There is one case when running sinking before stack layout could
-- be beneficial. Consider this:
--
--   L1:
--      x = y
--      call f() returns L2
--   L2: ...x...y...
--
-- Since both x and y are live across a call to f, they will be stored
-- on the stack during stack layout and restored after the call:
--
--   L1:
--      x = y
--      P64[Sp - 24] = L2
--      P64[Sp - 16] = x
--      P64[Sp - 8]  = y
--      Sp = Sp - 24
--      call f() returns L2
--   L2:
--      y = P64[Sp + 16]
--      x = P64[Sp + 8]
--      Sp = Sp + 24
--      ...x...y...
--
-- However, if we run sinking before stack layout we would propagate x
-- to its usage place (both x and y must be local register for this to
-- be possible - global registers cannot be floated past a call):
--
--   L1:
--      x = y
--      call f() returns L2
--   L2: ...y...y...
--
-- Thus making x dead at the call to f(). If we ran stack layout now
-- we would generate less stores and loads:
--
--   L1:
--      x = y
--      P64[Sp - 16] = L2
--      P64[Sp - 8]  = y
--      Sp = Sp - 16
--      call f() returns L2
--   L2:
--      y = P64[Sp + 8]
--      Sp = Sp + 16
--      ...y...y...
--
-- But since we don't see any benefits from running sinking before stack
-- layout, this situation probably doesn't arise too often in practice.
--

{- Note [inconsistent-pic-reg]

On x86/Darwin, PIC is implemented by inserting a sequence like

    call 1f
 1: popl %reg

at the proc entry point, and then referring to labels as offsets from
%reg.  If we don't split proc points, then we could have many entry
points in a proc that would need this sequence, and each entry point
would then get a different value for %reg.  If there are any join
points, then at the join point we don't have a consistent value for
%reg, so we don't know how to refer to labels.

Hence, on x86/Darwin, we have to split proc points, and then each proc
point will get its own PIC initialisation sequence.

This isn't an issue on x86/ELF, where the sequence is

    call 1f
 1: popl %reg
    addl $_GLOBAL_OFFSET_TABLE_+(.-1b), %reg

so %reg always has a consistent value: the address of
_GLOBAL_OFFSET_TABLE_, regardless of which entry point we arrived via.

-}

{- Note [unreachable blocks]

The control-flow optimiser sometimes leaves unreachable blocks behind
containing junk code.  These aren't necessarily a problem, but
removing them is good because it might save time in the native code
generator later.

-}

runUniqSM :: UniqSM a -> IO a
runUniqSM :: forall a. UniqSM a -> IO a
runUniqSM UniqSM a
m = do
  UniqSupply
us <- Char -> IO UniqSupply
mkSplitUniqSupply Char
'u'
  a -> IO a
forall (m :: * -> *) a. Monad m => a -> m a
return (UniqSupply -> UniqSM a -> a
forall a. UniqSupply -> UniqSM a -> a
initUs_ UniqSupply
us UniqSM a
m)


dumpGraph :: DynFlags -> DumpFlag -> String -> CmmGraph -> IO ()
dumpGraph :: DynFlags -> DumpFlag -> String -> CmmGraph -> IO ()
dumpGraph DynFlags
dflags DumpFlag
flag String
name CmmGraph
g = do
  Bool -> IO () -> IO ()
forall (f :: * -> *). Applicative f => Bool -> f () -> f ()
when (GeneralFlag -> DynFlags -> Bool
gopt GeneralFlag
Opt_DoCmmLinting DynFlags
dflags) (IO () -> IO ()) -> IO () -> IO ()
forall a b. (a -> b) -> a -> b
$ CmmGraph -> IO ()
do_lint CmmGraph
g
  DynFlags -> DumpFlag -> String -> DumpFormat -> SDoc -> IO ()
dumpWith DynFlags
dflags DumpFlag
flag String
name DumpFormat
FormatCMM (CmmGraph -> SDoc
forall a. Outputable a => a -> SDoc
ppr CmmGraph
g)
 where
  do_lint :: CmmGraph -> IO ()
do_lint CmmGraph
g = case DynFlags -> CmmGraph -> Maybe SDoc
cmmLintGraph DynFlags
dflags CmmGraph
g of
                 Just SDoc
err -> do { DynFlags -> SDoc -> IO ()
fatalErrorMsg DynFlags
dflags SDoc
err
                                ; DynFlags -> Int -> IO ()
ghcExit DynFlags
dflags Int
1
                                }
                 Maybe SDoc
Nothing  -> () -> IO ()
forall (m :: * -> *) a. Monad m => a -> m a
return ()

dumpWith :: DynFlags -> DumpFlag -> String -> DumpFormat -> SDoc -> IO ()
dumpWith :: DynFlags -> DumpFlag -> String -> DumpFormat -> SDoc -> IO ()
dumpWith DynFlags
dflags DumpFlag
flag String
txt DumpFormat
fmt SDoc
sdoc = do
  DynFlags -> DumpFlag -> String -> DumpFormat -> SDoc -> IO ()
dumpIfSet_dyn DynFlags
dflags DumpFlag
flag String
txt DumpFormat
fmt SDoc
sdoc
  Bool -> IO () -> IO ()
forall (f :: * -> *). Applicative f => Bool -> f () -> f ()
when (Bool -> Bool
not (DumpFlag -> DynFlags -> Bool
dopt DumpFlag
flag DynFlags
dflags)) (IO () -> IO ()) -> IO () -> IO ()
forall a b. (a -> b) -> a -> b
$
    -- If `-ddump-cmm-verbose -ddump-to-file` is specified,
    -- dump each Cmm pipeline stage output to a separate file.  #16930
    Bool -> IO () -> IO ()
forall (f :: * -> *). Applicative f => Bool -> f () -> f ()
when (DumpFlag -> DynFlags -> Bool
dopt DumpFlag
Opt_D_dump_cmm_verbose DynFlags
dflags)
      (IO () -> IO ()) -> IO () -> IO ()
forall a b. (a -> b) -> a -> b
$ DumpAction
dumpAction DynFlags
dflags (PrintUnqualified -> PprStyle
mkDumpStyle PrintUnqualified
alwaysQualify)
                   (DumpFlag -> DumpOptions
dumpOptionsFromFlag DumpFlag
flag) String
txt DumpFormat
fmt SDoc
sdoc
  DynFlags -> DumpFlag -> String -> DumpFormat -> SDoc -> IO ()
dumpIfSet_dyn DynFlags
dflags DumpFlag
Opt_D_dump_cmm_verbose_by_proc String
txt DumpFormat
fmt SDoc
sdoc