-- ----------------------------------------------------------------------------- -- -- (c) The University of Glasgow 1993-2004 -- -- This is the top-level module in the native code generator. -- -- ----------------------------------------------------------------------------- \begin{code} {-# OPTIONS -fno-warn-tabs #-} -- The above warning supression flag is a temporary kludge. -- While working on this module you are encouraged to remove it and -- detab the module (please do the detabbing in a separate patch). See -- http://hackage.haskell.org/trac/ghc/wiki/Commentary/CodingStyle#TabsvsSpaces -- for details module AsmCodeGen ( nativeCodeGen ) where #include "HsVersions.h" #include "nativeGen/NCG.h" import qualified X86.CodeGen import qualified X86.Regs import qualified X86.Instr import qualified X86.Ppr import qualified SPARC.CodeGen import qualified SPARC.Regs import qualified SPARC.Instr import qualified SPARC.Ppr import qualified SPARC.ShortcutJump import qualified SPARC.CodeGen.Expand import qualified PPC.CodeGen import qualified PPC.Cond import qualified PPC.Regs import qualified PPC.RegInfo import qualified PPC.Instr import qualified PPC.Ppr import RegAlloc.Liveness import qualified RegAlloc.Linear.Main as Linear import qualified GraphColor as Color import qualified RegAlloc.Graph.Main as Color import qualified RegAlloc.Graph.Stats as Color import qualified RegAlloc.Graph.TrivColorable as Color import TargetReg import Platform import Config import Instruction import PIC import Reg import NCGMonad import BlockId import CgUtils ( fixStgRegisters ) import OldCmm import CmmOpt ( cmmEliminateDeadBlocks, cmmMiniInline, cmmMachOpFold ) import OldPprCmm import CLabel import UniqFM import Unique ( Unique, getUnique ) import UniqSupply import DynFlags import StaticFlags import Util import BasicTypes ( Alignment ) import Digraph import Pretty (Doc) import qualified Pretty import BufWrite import Outputable import FastString import UniqSet import ErrUtils import Module -- DEBUGGING ONLY --import OrdList import Data.List import Data.Maybe import Control.Monad import System.IO {- The native-code generator has machine-independent and machine-dependent modules. This module ("AsmCodeGen") is the top-level machine-independent module. Before entering machine-dependent land, we do some machine-independent optimisations (defined below) on the 'CmmStmts's. We convert to the machine-specific 'Instr' datatype with 'cmmCodeGen', assuming an infinite supply of registers. We then use a machine-independent register allocator ('regAlloc') to rejoin reality. Obviously, 'regAlloc' has machine-specific helper functions (see about "RegAllocInfo" below). Finally, we order the basic blocks of the function so as to minimise the number of jumps between blocks, by utilising fallthrough wherever possible. The machine-dependent bits break down as follows: * ["MachRegs"] Everything about the target platform's machine registers (and immediate operands, and addresses, which tend to intermingle/interact with registers). * ["MachInstrs"] Includes the 'Instr' datatype (possibly should have a module of its own), plus a miscellany of other things (e.g., 'targetDoubleSize', 'smStablePtrTable', ...) * ["MachCodeGen"] is where 'Cmm' stuff turns into machine instructions. * ["PprMach"] 'pprInstr' turns an 'Instr' into text (well, really a 'Doc'). * ["RegAllocInfo"] In the register allocator, we manipulate 'MRegsState's, which are 'BitSet's, one bit per machine register. When we want to say something about a specific machine register (e.g., ``it gets clobbered by this instruction''), we set/unset its bit. Obviously, we do this 'BitSet' thing for efficiency reasons. The 'RegAllocInfo' module collects together the machine-specific info needed to do register allocation. * ["RegisterAlloc"] The (machine-independent) register allocator. -} -- ----------------------------------------------------------------------------- -- Top-level of the native codegen data NcgImpl statics instr jumpDest = NcgImpl { cmmTopCodeGen :: RawCmmDecl -> NatM [NatCmmDecl statics instr], generateJumpTableForInstr :: instr -> Maybe (NatCmmDecl statics instr), getJumpDestBlockId :: jumpDest -> Maybe BlockId, canShortcut :: instr -> Maybe jumpDest, shortcutStatics :: (BlockId -> Maybe jumpDest) -> statics -> statics, shortcutJump :: (BlockId -> Maybe jumpDest) -> instr -> instr, pprNatCmmDecl :: Platform -> NatCmmDecl statics instr -> Doc, maxSpillSlots :: Int, allocatableRegs :: [RealReg], ncg_x86fp_kludge :: [NatCmmDecl statics instr] -> [NatCmmDecl statics instr], ncgExpandTop :: [NatCmmDecl statics instr] -> [NatCmmDecl statics instr], ncgMakeFarBranches :: [NatBasicBlock instr] -> [NatBasicBlock instr] } -------------------- nativeCodeGen :: DynFlags -> Handle -> UniqSupply -> [RawCmmGroup] -> IO () nativeCodeGen dflags h us cmms = let platform = targetPlatform dflags nCG' :: (PlatformOutputable statics, PlatformOutputable instr, Instruction instr) => NcgImpl statics instr jumpDest -> IO () nCG' ncgImpl = nativeCodeGen' dflags ncgImpl h us cmms x86NcgImpl = NcgImpl { cmmTopCodeGen = X86.CodeGen.cmmTopCodeGen ,generateJumpTableForInstr = X86.CodeGen.generateJumpTableForInstr ,getJumpDestBlockId = X86.Instr.getJumpDestBlockId ,canShortcut = X86.Instr.canShortcut ,shortcutStatics = X86.Instr.shortcutStatics ,shortcutJump = X86.Instr.shortcutJump ,pprNatCmmDecl = X86.Ppr.pprNatCmmDecl ,maxSpillSlots = X86.Instr.maxSpillSlots (target32Bit platform) ,allocatableRegs = X86.Regs.allocatableRegs ,ncg_x86fp_kludge = id ,ncgExpandTop = id ,ncgMakeFarBranches = id } in case platformArch platform of ArchX86 -> nCG' (x86NcgImpl { ncg_x86fp_kludge = map x86fp_kludge }) ArchX86_64 -> nCG' x86NcgImpl ArchPPC -> nCG' $ NcgImpl { cmmTopCodeGen = PPC.CodeGen.cmmTopCodeGen ,generateJumpTableForInstr = PPC.CodeGen.generateJumpTableForInstr ,getJumpDestBlockId = PPC.RegInfo.getJumpDestBlockId ,canShortcut = PPC.RegInfo.canShortcut ,shortcutStatics = PPC.RegInfo.shortcutStatics ,shortcutJump = PPC.RegInfo.shortcutJump ,pprNatCmmDecl = PPC.Ppr.pprNatCmmDecl ,maxSpillSlots = PPC.Instr.maxSpillSlots ,allocatableRegs = PPC.Regs.allocatableRegs ,ncg_x86fp_kludge = id ,ncgExpandTop = id ,ncgMakeFarBranches = makeFarBranches } ArchSPARC -> nCG' $ NcgImpl { cmmTopCodeGen = SPARC.CodeGen.cmmTopCodeGen ,generateJumpTableForInstr = SPARC.CodeGen.generateJumpTableForInstr ,getJumpDestBlockId = SPARC.ShortcutJump.getJumpDestBlockId ,canShortcut = SPARC.ShortcutJump.canShortcut ,shortcutStatics = SPARC.ShortcutJump.shortcutStatics ,shortcutJump = SPARC.ShortcutJump.shortcutJump ,pprNatCmmDecl = SPARC.Ppr.pprNatCmmDecl ,maxSpillSlots = SPARC.Instr.maxSpillSlots ,allocatableRegs = SPARC.Regs.allocatableRegs ,ncg_x86fp_kludge = id ,ncgExpandTop = map SPARC.CodeGen.Expand.expandTop ,ncgMakeFarBranches = id } ArchARM _ _ -> panic "nativeCodeGen: No NCG for ARM" ArchPPC_64 -> panic "nativeCodeGen: No NCG for PPC 64" ArchUnknown -> panic "nativeCodeGen: No NCG for unknown arch" nativeCodeGen' :: (PlatformOutputable statics, PlatformOutputable instr, Instruction instr) => DynFlags -> NcgImpl statics instr jumpDest -> Handle -> UniqSupply -> [RawCmmGroup] -> IO () nativeCodeGen' dflags ncgImpl h us cmms = do let platform = targetPlatform dflags split_cmms = concat $ map add_split cmms -- BufHandle is a performance hack. We could hide it inside -- Pretty if it weren't for the fact that we do lots of little -- printDocs here (in order to do codegen in constant space). bufh <- newBufHandle h (imports, prof) <- cmmNativeGens dflags ncgImpl bufh us split_cmms [] [] 0 bFlush bufh let (native, colorStats, linearStats) = unzip3 prof -- dump native code dumpIfSet_dyn dflags Opt_D_dump_asm "Asm code" (vcat $ map (docToSDoc . pprNatCmmDecl ncgImpl platform) $ concat native) -- dump global NCG stats for graph coloring allocator (case concat $ catMaybes colorStats of [] -> return () stats -> do -- build the global register conflict graph let graphGlobal = foldl Color.union Color.initGraph $ [ Color.raGraph stat | stat@Color.RegAllocStatsStart{} <- stats] dumpSDoc dflags Opt_D_dump_asm_stats "NCG stats" $ Color.pprStats stats graphGlobal dumpIfSet_dyn dflags Opt_D_dump_asm_conflicts "Register conflict graph" $ Color.dotGraph (targetRegDotColor platform) (Color.trivColorable platform (targetVirtualRegSqueeze platform) (targetRealRegSqueeze platform)) $ graphGlobal) -- dump global NCG stats for linear allocator (case concat $ catMaybes linearStats of [] -> return () stats -> dumpSDoc dflags Opt_D_dump_asm_stats "NCG stats" $ Linear.pprStats (concat native) stats) -- write out the imports Pretty.printDoc Pretty.LeftMode h $ makeImportsDoc dflags (concat imports) return () where add_split tops | dopt Opt_SplitObjs dflags = split_marker : tops | otherwise = tops split_marker = CmmProc Nothing mkSplitMarkerLabel (ListGraph []) -- | Do native code generation on all these cmms. -- cmmNativeGens :: (PlatformOutputable statics, PlatformOutputable instr, Instruction instr) => DynFlags -> NcgImpl statics instr jumpDest -> BufHandle -> UniqSupply -> [RawCmmDecl] -> [[CLabel]] -> [ ([NatCmmDecl statics instr], Maybe [Color.RegAllocStats statics instr], Maybe [Linear.RegAllocStats]) ] -> Int -> IO ( [[CLabel]], [([NatCmmDecl statics instr], Maybe [Color.RegAllocStats statics instr], Maybe [Linear.RegAllocStats])] ) cmmNativeGens _ _ _ _ [] impAcc profAcc _ = return (reverse impAcc, reverse profAcc) cmmNativeGens dflags ncgImpl h us (cmm : cmms) impAcc profAcc count = do let platform = targetPlatform dflags (us', native, imports, colorStats, linearStats) <- {-# SCC "cmmNativeGen" #-} cmmNativeGen dflags ncgImpl us cmm count {-# SCC "pprNativeCode" #-} Pretty.bufLeftRender h $ Pretty.vcat $ map (pprNatCmmDecl ncgImpl platform) native -- carefully evaluate this strictly. Binding it with 'let' -- and then using 'seq' doesn't work, because the let -- apparently gets inlined first. lsPprNative <- return $! if dopt Opt_D_dump_asm dflags || dopt Opt_D_dump_asm_stats dflags then native else [] count' <- return $! count + 1; -- force evaulation all this stuff to avoid space leaks {-# SCC "seqString" #-} seqString (showSDoc $ vcat $ map (pprPlatform platform) imports) `seq` return () cmmNativeGens dflags ncgImpl h us' cmms (imports : impAcc) ((lsPprNative, colorStats, linearStats) : profAcc) count' where seqString [] = () seqString (x:xs) = x `seq` seqString xs `seq` () -- | Complete native code generation phase for a single top-level chunk of Cmm. -- Dumping the output of each stage along the way. -- Global conflict graph and NGC stats cmmNativeGen :: (PlatformOutputable statics, PlatformOutputable instr, Instruction instr) => DynFlags -> NcgImpl statics instr jumpDest -> UniqSupply -> RawCmmDecl -- ^ the cmm to generate code for -> Int -- ^ sequence number of this top thing -> IO ( UniqSupply , [NatCmmDecl statics instr] -- native code , [CLabel] -- things imported by this cmm , Maybe [Color.RegAllocStats statics instr] -- stats for the coloring register allocator , Maybe [Linear.RegAllocStats]) -- stats for the linear register allocators cmmNativeGen dflags ncgImpl us cmm count = do let platform = targetPlatform dflags -- rewrite assignments to global regs let fixed_cmm = {-# SCC "fixStgRegisters" #-} fixStgRegisters cmm -- cmm to cmm optimisations let (opt_cmm, imports) = {-# SCC "cmmToCmm" #-} cmmToCmm dflags fixed_cmm dumpIfSet_dyn dflags Opt_D_dump_opt_cmm "Optimised Cmm" (pprCmmGroup platform [opt_cmm]) -- generate native code from cmm let ((native, lastMinuteImports), usGen) = {-# SCC "genMachCode" #-} initUs us $ genMachCode dflags (cmmTopCodeGen ncgImpl) opt_cmm dumpIfSet_dyn dflags Opt_D_dump_asm_native "Native code" (vcat $ map (docToSDoc . pprNatCmmDecl ncgImpl platform) native) -- tag instructions with register liveness information let (withLiveness, usLive) = {-# SCC "regLiveness" #-} initUs usGen $ mapUs (regLiveness platform) $ map natCmmTopToLive native dumpIfSet_dyn dflags Opt_D_dump_asm_liveness "Liveness annotations added" (vcat $ map (pprPlatform platform) withLiveness) -- allocate registers (alloced, usAlloc, ppr_raStatsColor, ppr_raStatsLinear) <- if ( dopt Opt_RegsGraph dflags || dopt Opt_RegsIterative dflags) then do -- the regs usable for allocation let (alloc_regs :: UniqFM (UniqSet RealReg)) = foldr (\r -> plusUFM_C unionUniqSets $ unitUFM (targetClassOfRealReg platform r) (unitUniqSet r)) emptyUFM $ allocatableRegs ncgImpl -- do the graph coloring register allocation let ((alloced, regAllocStats), usAlloc) = {-# SCC "RegAlloc" #-} initUs usLive $ Color.regAlloc dflags alloc_regs (mkUniqSet [0 .. maxSpillSlots ncgImpl]) withLiveness -- dump out what happened during register allocation dumpIfSet_dyn dflags Opt_D_dump_asm_regalloc "Registers allocated" (vcat $ map (docToSDoc . pprNatCmmDecl ncgImpl platform) alloced) dumpIfSet_dyn dflags Opt_D_dump_asm_regalloc_stages "Build/spill stages" (vcat $ map (\(stage, stats) -> text "# --------------------------" $$ text "# cmm " <> int count <> text " Stage " <> int stage $$ pprPlatform platform stats) $ zip [0..] regAllocStats) let mPprStats = if dopt Opt_D_dump_asm_stats dflags then Just regAllocStats else Nothing -- force evaluation of the Maybe to avoid space leak mPprStats `seq` return () return ( alloced, usAlloc , mPprStats , Nothing) else do -- do linear register allocation let ((alloced, regAllocStats), usAlloc) = {-# SCC "RegAlloc" #-} initUs usLive $ liftM unzip $ mapUs (Linear.regAlloc dflags) withLiveness dumpIfSet_dyn dflags Opt_D_dump_asm_regalloc "Registers allocated" (vcat $ map (docToSDoc . pprNatCmmDecl ncgImpl platform) alloced) let mPprStats = if dopt Opt_D_dump_asm_stats dflags then Just (catMaybes regAllocStats) else Nothing -- force evaluation of the Maybe to avoid space leak mPprStats `seq` return () return ( alloced, usAlloc , Nothing , mPprStats) ---- x86fp_kludge. This pass inserts ffree instructions to clear ---- the FPU stack on x86. The x86 ABI requires that the FPU stack ---- is clear, and library functions can return odd results if it ---- isn't. ---- ---- NB. must happen before shortcutBranches, because that ---- generates JXX_GBLs which we can't fix up in x86fp_kludge. let kludged = {-# SCC "x86fp_kludge" #-} ncg_x86fp_kludge ncgImpl alloced ---- generate jump tables let tabled = {-# SCC "generateJumpTables" #-} generateJumpTables ncgImpl kludged ---- shortcut branches let shorted = {-# SCC "shortcutBranches" #-} shortcutBranches dflags ncgImpl tabled ---- sequence blocks let sequenced = {-# SCC "sequenceBlocks" #-} map (sequenceTop ncgImpl) shorted ---- expansion of SPARC synthetic instrs let expanded = {-# SCC "sparc_expand" #-} ncgExpandTop ncgImpl sequenced dumpIfSet_dyn dflags Opt_D_dump_asm_expanded "Synthetic instructions expanded" (vcat $ map (docToSDoc . pprNatCmmDecl ncgImpl platform) expanded) return ( usAlloc , expanded , lastMinuteImports ++ imports , ppr_raStatsColor , ppr_raStatsLinear) x86fp_kludge :: NatCmmDecl (Alignment, CmmStatics) X86.Instr.Instr -> NatCmmDecl (Alignment, CmmStatics) X86.Instr.Instr x86fp_kludge top@(CmmData _ _) = top x86fp_kludge (CmmProc info lbl (ListGraph code)) = CmmProc info lbl (ListGraph $ X86.Instr.i386_insert_ffrees code) -- | Build a doc for all the imports. -- makeImportsDoc :: DynFlags -> [CLabel] -> Pretty.Doc makeImportsDoc dflags imports = dyld_stubs imports Pretty.$$ -- On recent versions of Darwin, the linker supports -- dead-stripping of code and data on a per-symbol basis. -- There's a hack to make this work in PprMach.pprNatCmmDecl. (if platformHasSubsectionsViaSymbols (targetPlatform dflags) then Pretty.text ".subsections_via_symbols" else Pretty.empty) Pretty.$$ -- On recent GNU ELF systems one can mark an object file -- as not requiring an executable stack. If all objects -- linked into a program have this note then the program -- will not use an executable stack, which is good for -- security. GHC generated code does not need an executable -- stack so add the note in: (if platformHasGnuNonexecStack (targetPlatform dflags) then Pretty.text ".section .note.GNU-stack,\"\",@progbits" else Pretty.empty) Pretty.$$ -- And just because every other compiler does, lets stick in -- an identifier directive: .ident "GHC x.y.z" (if platformHasIdentDirective (targetPlatform dflags) then let compilerIdent = Pretty.text "GHC" Pretty.<+> Pretty.text cProjectVersion in Pretty.text ".ident" Pretty.<+> Pretty.doubleQuotes compilerIdent else Pretty.empty) where -- Generate "symbol stubs" for all external symbols that might -- come from a dynamic library. dyld_stubs :: [CLabel] -> Pretty.Doc {- dyld_stubs imps = Pretty.vcat $ map pprDyldSymbolStub $ map head $ group $ sort imps-} platform = targetPlatform dflags arch = platformArch platform os = platformOS platform -- (Hack) sometimes two Labels pretty-print the same, but have -- different uniques; so we compare their text versions... dyld_stubs imps | needImportedSymbols arch os = Pretty.vcat $ (pprGotDeclaration arch os :) $ map ( pprImportedSymbol platform . fst . head) $ groupBy (\(_,a) (_,b) -> a == b) $ sortBy (\(_,a) (_,b) -> compare a b) $ map doPpr $ imps | otherwise = Pretty.empty doPpr lbl = (lbl, renderWithStyle (pprCLabel platform lbl) astyle) astyle = mkCodeStyle AsmStyle -- ----------------------------------------------------------------------------- -- Sequencing the basic blocks -- Cmm BasicBlocks are self-contained entities: they always end in a -- jump, either non-local or to another basic block in the same proc. -- In this phase, we attempt to place the basic blocks in a sequence -- such that as many of the local jumps as possible turn into -- fallthroughs. sequenceTop :: Instruction instr => NcgImpl statics instr jumpDest -> NatCmmDecl statics instr -> NatCmmDecl statics instr sequenceTop _ top@(CmmData _ _) = top sequenceTop ncgImpl (CmmProc info lbl (ListGraph blocks)) = CmmProc info lbl (ListGraph $ ncgMakeFarBranches ncgImpl $ sequenceBlocks blocks) -- The algorithm is very simple (and stupid): we make a graph out of -- the blocks where there is an edge from one block to another iff the -- first block ends by jumping to the second. Then we topologically -- sort this graph. Then traverse the list: for each block, we first -- output the block, then if it has an out edge, we move the -- destination of the out edge to the front of the list, and continue. -- FYI, the classic layout for basic blocks uses postorder DFS; this -- algorithm is implemented in Hoopl. sequenceBlocks :: Instruction instr => [NatBasicBlock instr] -> [NatBasicBlock instr] sequenceBlocks [] = [] sequenceBlocks (entry:blocks) = seqBlocks (mkNode entry : reverse (flattenSCCs (sccBlocks blocks))) -- the first block is the entry point ==> it must remain at the start. sccBlocks :: Instruction instr => [NatBasicBlock instr] -> [SCC ( NatBasicBlock instr , Unique , [Unique])] sccBlocks blocks = stronglyConnCompFromEdgedVerticesR (map mkNode blocks) -- we're only interested in the last instruction of -- the block, and only if it has a single destination. getOutEdges :: Instruction instr => [instr] -> [Unique] getOutEdges instrs = case jumpDestsOfInstr (last instrs) of [one] -> [getUnique one] _many -> [] mkNode :: (Instruction t) => GenBasicBlock t -> (GenBasicBlock t, Unique, [Unique]) mkNode block@(BasicBlock id instrs) = (block, getUnique id, getOutEdges instrs) seqBlocks :: (Eq t) => [(GenBasicBlock t1, t, [t])] -> [GenBasicBlock t1] seqBlocks [] = [] seqBlocks ((block,_,[]) : rest) = block : seqBlocks rest seqBlocks ((block@(BasicBlock id instrs),_,[next]) : rest) | can_fallthrough = BasicBlock id (init instrs) : seqBlocks rest' | otherwise = block : seqBlocks rest' where (can_fallthrough, rest') = reorder next [] rest -- TODO: we should do a better job for cycles; try to maximise the -- fallthroughs within a loop. seqBlocks _ = panic "AsmCodegen:seqBlocks" reorder :: (Eq a) => a -> [(t, a, t1)] -> [(t, a, t1)] -> (Bool, [(t, a, t1)]) reorder _ accum [] = (False, reverse accum) reorder id accum (b@(block,id',out) : rest) | id == id' = (True, (block,id,out) : reverse accum ++ rest) | otherwise = reorder id (b:accum) rest -- ----------------------------------------------------------------------------- -- Making far branches -- Conditional branches on PowerPC are limited to +-32KB; if our Procs get too -- big, we have to work around this limitation. makeFarBranches :: [NatBasicBlock PPC.Instr.Instr] -> [NatBasicBlock PPC.Instr.Instr] makeFarBranches blocks | last blockAddresses < nearLimit = blocks | otherwise = zipWith handleBlock blockAddresses blocks where blockAddresses = scanl (+) 0 $ map blockLen blocks blockLen (BasicBlock _ instrs) = length instrs handleBlock addr (BasicBlock id instrs) = BasicBlock id (zipWith makeFar [addr..] instrs) makeFar _ (PPC.Instr.BCC PPC.Cond.ALWAYS tgt) = PPC.Instr.BCC PPC.Cond.ALWAYS tgt makeFar addr (PPC.Instr.BCC cond tgt) | abs (addr - targetAddr) >= nearLimit = PPC.Instr.BCCFAR cond tgt | otherwise = PPC.Instr.BCC cond tgt where Just targetAddr = lookupUFM blockAddressMap tgt makeFar _ other = other nearLimit = 7000 -- 8192 instructions are allowed; let's keep some -- distance, as we have a few pseudo-insns that are -- pretty-printed as multiple instructions, -- and it's just not worth the effort to calculate -- things exactly blockAddressMap = listToUFM $ zip (map blockId blocks) blockAddresses -- ----------------------------------------------------------------------------- -- Generate jump tables -- Analyzes all native code and generates data sections for all jump -- table instructions. generateJumpTables :: NcgImpl statics instr jumpDest -> [NatCmmDecl statics instr] -> [NatCmmDecl statics instr] generateJumpTables ncgImpl xs = concatMap f xs where f p@(CmmProc _ _ (ListGraph xs)) = p : concatMap g xs f p = [p] g (BasicBlock _ xs) = catMaybes (map (generateJumpTableForInstr ncgImpl) xs) -- ----------------------------------------------------------------------------- -- Shortcut branches shortcutBranches :: DynFlags -> NcgImpl statics instr jumpDest -> [NatCmmDecl statics instr] -> [NatCmmDecl statics instr] shortcutBranches dflags ncgImpl tops | optLevel dflags < 1 = tops -- only with -O or higher | otherwise = map (apply_mapping ncgImpl mapping) tops' where (tops', mappings) = mapAndUnzip (build_mapping ncgImpl) tops mapping = foldr plusUFM emptyUFM mappings build_mapping :: NcgImpl statics instr jumpDest -> GenCmmDecl d t (ListGraph instr) -> (GenCmmDecl d t (ListGraph instr), UniqFM jumpDest) build_mapping _ top@(CmmData _ _) = (top, emptyUFM) build_mapping _ (CmmProc info lbl (ListGraph [])) = (CmmProc info lbl (ListGraph []), emptyUFM) build_mapping ncgImpl (CmmProc info lbl (ListGraph (head:blocks))) = (CmmProc info lbl (ListGraph (head:others)), mapping) -- drop the shorted blocks, but don't ever drop the first one, -- because it is pointed to by a global label. where -- find all the blocks that just consist of a jump that can be -- shorted. -- Don't completely eliminate loops here -- that can leave a dangling jump! (_, shortcut_blocks, others) = foldl split (emptyBlockSet, [], []) blocks split (s, shortcut_blocks, others) b@(BasicBlock id [insn]) | Just jd <- canShortcut ncgImpl insn, Just dest <- getJumpDestBlockId ncgImpl jd, (setMember dest s) || dest == id -- loop checks = (s, shortcut_blocks, b : others) split (s, shortcut_blocks, others) (BasicBlock id [insn]) | Just dest <- canShortcut ncgImpl insn = (setInsert id s, (id,dest) : shortcut_blocks, others) split (s, shortcut_blocks, others) other = (s, shortcut_blocks, other : others) -- build a mapping from BlockId to JumpDest for shorting branches mapping = foldl add emptyUFM shortcut_blocks add ufm (id,dest) = addToUFM ufm id dest apply_mapping :: NcgImpl statics instr jumpDest -> UniqFM jumpDest -> GenCmmDecl statics h (ListGraph instr) -> GenCmmDecl statics h (ListGraph instr) apply_mapping ncgImpl ufm (CmmData sec statics) = CmmData sec (shortcutStatics ncgImpl (lookupUFM ufm) statics) apply_mapping ncgImpl ufm (CmmProc info lbl (ListGraph blocks)) = CmmProc info lbl (ListGraph $ map short_bb blocks) where short_bb (BasicBlock id insns) = BasicBlock id $! map short_insn insns short_insn i = shortcutJump ncgImpl (lookupUFM ufm) i -- shortcutJump should apply the mapping repeatedly, -- just in case we can short multiple branches. -- ----------------------------------------------------------------------------- -- Instruction selection -- Native code instruction selection for a chunk of stix code. For -- this part of the computation, we switch from the UniqSM monad to -- the NatM monad. The latter carries not only a Unique, but also an -- Int denoting the current C stack pointer offset in the generated -- code; this is needed for creating correct spill offsets on -- architectures which don't offer, or for which it would be -- prohibitively expensive to employ, a frame pointer register. Viz, -- x86. -- The offset is measured in bytes, and indicates the difference -- between the current (simulated) C stack-ptr and the value it was at -- the beginning of the block. For stacks which grow down, this value -- should be either zero or negative. -- Switching between the two monads whilst carrying along the same -- Unique supply breaks abstraction. Is that bad? genMachCode :: DynFlags -> (RawCmmDecl -> NatM [NatCmmDecl statics instr]) -> RawCmmDecl -> UniqSM ( [NatCmmDecl statics instr] , [CLabel]) genMachCode dflags cmmTopCodeGen cmm_top = do { initial_us <- getUs ; let initial_st = mkNatM_State initial_us 0 dflags (new_tops, final_st) = initNat initial_st (cmmTopCodeGen cmm_top) final_delta = natm_delta final_st final_imports = natm_imports final_st ; if final_delta == 0 then return (new_tops, final_imports) else pprPanic "genMachCode: nonzero final delta" (int final_delta) } -- ----------------------------------------------------------------------------- -- Generic Cmm optimiser {- Here we do: (a) Constant folding (b) Simple inlining: a temporary which is assigned to and then used, once, can be shorted. (c) Position independent code and dynamic linking (i) introduce the appropriate indirections and position independent refs (ii) compile a list of imported symbols (d) Some arch-specific optimizations (a) and (b) will be moving to the new Hoopl pipeline, however, (c) and (d) are only needed by the native backend and will continue to live here. Ideas for other things we could do (put these in Hoopl please!): - shortcut jumps-to-jumps - simple CSE: if an expr is assigned to a temp, then replace later occs of that expr with the temp, until the expr is no longer valid (can push through temp assignments, and certain assigns to mem...) -} cmmToCmm :: DynFlags -> RawCmmDecl -> (RawCmmDecl, [CLabel]) cmmToCmm _ top@(CmmData _ _) = (top, []) cmmToCmm dflags (CmmProc info lbl (ListGraph blocks)) = runCmmOpt dflags $ do let platform = targetPlatform dflags blocks' <- mapM cmmBlockConFold (cmmMiniInline platform (cmmEliminateDeadBlocks blocks)) return $ CmmProc info lbl (ListGraph blocks') newtype CmmOptM a = CmmOptM (([CLabel], DynFlags) -> (# a, [CLabel] #)) instance Monad CmmOptM where return x = CmmOptM $ \(imports, _) -> (# x,imports #) (CmmOptM f) >>= g = CmmOptM $ \(imports, dflags) -> case f (imports, dflags) of (# x, imports' #) -> case g x of CmmOptM g' -> g' (imports', dflags) addImportCmmOpt :: CLabel -> CmmOptM () addImportCmmOpt lbl = CmmOptM $ \(imports, _dflags) -> (# (), lbl:imports #) instance HasDynFlags CmmOptM where getDynFlags = CmmOptM $ \(imports, dflags) -> (# dflags, imports #) runCmmOpt :: DynFlags -> CmmOptM a -> (a, [CLabel]) runCmmOpt dflags (CmmOptM f) = case f ([], dflags) of (# result, imports #) -> (result, imports) cmmBlockConFold :: CmmBasicBlock -> CmmOptM CmmBasicBlock cmmBlockConFold (BasicBlock id stmts) = do stmts' <- mapM cmmStmtConFold stmts return $ BasicBlock id stmts' -- This does three optimizations, but they're very quick to check, so we don't -- bother turning them off even when the Hoopl code is active. Since -- this is on the old Cmm representation, we can't reuse the code either: -- * reg = reg --> nop -- * if 0 then jump --> nop -- * if 1 then jump --> jump -- We might be tempted to skip this step entirely of not opt_PIC, but -- there is some PowerPC code for the non-PIC case, which would also -- have to be separated. cmmStmtConFold :: CmmStmt -> CmmOptM CmmStmt cmmStmtConFold stmt = case stmt of CmmAssign reg src -> do src' <- cmmExprConFold DataReference src return $ case src' of CmmReg reg' | reg == reg' -> CmmNop new_src -> CmmAssign reg new_src CmmStore addr src -> do addr' <- cmmExprConFold DataReference addr src' <- cmmExprConFold DataReference src return $ CmmStore addr' src' CmmJump addr live -> do addr' <- cmmExprConFold JumpReference addr return $ CmmJump addr' live CmmCall target regs args returns -> do target' <- case target of CmmCallee e conv -> do e' <- cmmExprConFold CallReference e return $ CmmCallee e' conv other -> return other args' <- mapM (\(CmmHinted arg hint) -> do arg' <- cmmExprConFold DataReference arg return (CmmHinted arg' hint)) args return $ CmmCall target' regs args' returns CmmCondBranch test dest -> do test' <- cmmExprConFold DataReference test dflags <- getDynFlags let platform = targetPlatform dflags return $ case test' of CmmLit (CmmInt 0 _) -> CmmComment (mkFastString ("deleted: " ++ showSDoc (pprStmt platform stmt))) CmmLit (CmmInt _ _) -> CmmBranch dest _other -> CmmCondBranch test' dest CmmSwitch expr ids -> do expr' <- cmmExprConFold DataReference expr return $ CmmSwitch expr' ids other -> return other cmmExprConFold :: ReferenceKind -> CmmExpr -> CmmOptM CmmExpr cmmExprConFold referenceKind expr = do dflags <- getDynFlags -- Skip constant folding if new code generator is running -- (this optimization is done in Hoopl) let expr' = if dopt Opt_TryNewCodeGen dflags then expr else cmmExprCon (targetPlatform dflags) expr cmmExprNative referenceKind expr' cmmExprCon :: Platform -> CmmExpr -> CmmExpr cmmExprCon platform (CmmLoad addr rep) = CmmLoad (cmmExprCon platform addr) rep cmmExprCon platform (CmmMachOp mop args) = cmmMachOpFold platform mop (map (cmmExprCon platform) args) cmmExprCon _ other = other -- handles both PIC and non-PIC cases... a very strange mixture -- of things to do. cmmExprNative :: ReferenceKind -> CmmExpr -> CmmOptM CmmExpr cmmExprNative referenceKind expr = do dflags <- getDynFlags let platform = targetPlatform dflags arch = platformArch platform case expr of CmmLoad addr rep -> do addr' <- cmmExprNative DataReference addr return $ CmmLoad addr' rep CmmMachOp mop args -> do args' <- mapM (cmmExprNative DataReference) args return $ CmmMachOp mop args' CmmLit (CmmLabel lbl) -> do cmmMakeDynamicReference dflags addImportCmmOpt referenceKind lbl CmmLit (CmmLabelOff lbl off) -> do dynRef <- cmmMakeDynamicReference dflags addImportCmmOpt referenceKind lbl -- need to optimize here, since it's late return $ cmmMachOpFold platform (MO_Add wordWidth) [ dynRef, (CmmLit $ CmmInt (fromIntegral off) wordWidth) ] -- On powerpc (non-PIC), it's easier to jump directly to a label than -- to use the register table, so we replace these registers -- with the corresponding labels: CmmReg (CmmGlobal EagerBlackholeInfo) | arch == ArchPPC && not opt_PIC -> cmmExprNative referenceKind $ CmmLit (CmmLabel (mkCmmCodeLabel rtsPackageId (fsLit "__stg_EAGER_BLACKHOLE_info"))) CmmReg (CmmGlobal GCEnter1) | arch == ArchPPC && not opt_PIC -> cmmExprNative referenceKind $ CmmLit (CmmLabel (mkCmmCodeLabel rtsPackageId (fsLit "__stg_gc_enter_1"))) CmmReg (CmmGlobal GCFun) | arch == ArchPPC && not opt_PIC -> cmmExprNative referenceKind $ CmmLit (CmmLabel (mkCmmCodeLabel rtsPackageId (fsLit "__stg_gc_fun"))) other -> return other \end{code}