diff --git a/binary.cabal b/binary.cabal
--- a/binary.cabal
+++ b/binary.cabal
@@ -1,5 +1,5 @@
 name:            binary
-version:         0.4.5
+version:         0.5
 license:         BSD3
 license-file:    LICENSE
 author:          Lennart Kolmodin <kolmodin@dtek.chalmers.se>
@@ -52,13 +52,7 @@
   extensions:      CPP,
                    FlexibleContexts
 
-  ghc-options:     -O2
-                   -Wall
-                   -fspec-constr
-                   -funbox-strict-fields 
-                   -fdicts-cheap
-                   -fliberate-case-threshold=1000
-                   -fmax-simplifier-iterations10
+  ghc-options:     -O2 -Wall -fliberate-case-threshold=1000
 
 --  if impl(ghc < 6.5)
 --    ghc-options:   -fallow-undecidable-instances
diff --git a/docs/hcar/binary-Lb.tex b/docs/hcar/binary-Lb.tex
new file mode 100644
--- /dev/null
+++ b/docs/hcar/binary-Lb.tex
@@ -0,0 +1,48 @@
+\begin{hcarentry}{binary}
+\label{binary}
+\report{Lennart Kolmodin}
+\status{active}
+\participants{Duncan Coutts, Don Stewart, Binary Strike Team}
+\makeheader
+
+The Binary Strike Team is pleased to announce yet a release of a new,
+pure, efficient binary serialisation library.
+
+The `binary' package provides efficient serialisation of Haskell values
+to and from lazy ByteStrings. ByteStrings constructed this way may then
+be written to disk, written to the network, or further processed (e.g.
+stored in memory directly, or compressed in memory with zlib or bzlib).
+
+The binary library has been heavily tuned for performance, particularly for
+writing speed. Throughput of up to 160M/s has been achieved in practice, and
+in general speed is on par or better than NewBinary, with the advantage of a
+pure interface. Efforts are underway to improve performance still further.
+Plans are also taking shape for a parser combinator library on top of
+binary, for bit parsing and foreign structure parsing (e.g. network
+protocols).
+
+Data.Derive~\cref{derive} has support for automatically generating Binary
+instances, allowing to read and write your data structures with little fuzz.
+
+Binary was developed by a team of 8 during the Haskell Hackathon in Oxford
+2007, and since then has about 15 people contributed code and many more
+given feedback and cheerleading on \verb|#haskell|.
+
+The package is cabalized and available through Hackage~\cref{hackagedb}.
+% to editors: ref. to cabal?
+
+\FurtherReading
+\begin{compactitem}
+\item Homepage
+
+  \url{http://code.haskell.org/binary/}
+\item Hackage
+
+  \url{http://hackage.haskell.org/cgi-bin/hackage-scripts/package/binary}
+\item Development version
+
+  \texttt{darcs get --partial}
+
+  \url{http://code.haskell.org/binary}
+\end{compactitem}
+\end{hcarentry}
diff --git a/src/Data/Binary.hs b/src/Data/Binary.hs
--- a/src/Data/Binary.hs
+++ b/src/Data/Binary.hs
@@ -651,9 +651,13 @@
 --
 
 instance (Binary e) => Binary (Seq.Seq e) where
-    -- any better way to do this?
-    put = put . Fold.toList
-    get = fmap Seq.fromList get
+    put s = put (Seq.length s) >> Fold.mapM_ put s
+    get = do n <- get :: Get Int
+             rep Seq.empty n get
+      where rep xs 0 _ = return $! xs
+            rep xs n g = xs `seq` n `seq` do
+                           x <- g
+                           rep (xs Seq.|> x) (n-1) g
 
 #endif
 
diff --git a/src/Data/Binary/Put.hs b/src/Data/Binary/Put.hs
--- a/src/Data/Binary/Put.hs
+++ b/src/Data/Binary/Put.hs
@@ -18,6 +18,9 @@
       Put
     , PutM(..)
     , runPut
+    , runPutM
+    , putBuilder
+    , execPut
 
     -- * Flushing the implicit parse state
     , flush
@@ -106,10 +109,24 @@
 tell b = Put $ PairS () b
 {-# INLINE tell #-}
 
+putBuilder :: Builder -> Put
+putBuilder = tell
+{-# INLINE putBuilder #-}
+
+-- | Run the 'Put' monad
+execPut :: PutM a -> Builder
+execPut = sndS . unPut
+{-# INLINE execPut #-}
+
 -- | Run the 'Put' monad with a serialiser
 runPut :: Put -> L.ByteString
 runPut = toLazyByteString . sndS . unPut
 {-# INLINE runPut #-}
+
+-- | Run the 'Put' monad with a serialiser and get its result
+runPutM :: PutM a -> (a, L.ByteString)
+runPutM (Put (PairS f s)) = (f, toLazyByteString s)
+{-# INLINE runPutM #-}
 
 ------------------------------------------------------------------------
 
diff --git a/tests/Benchmark.hs b/tests/Benchmark.hs
new file mode 100644
--- /dev/null
+++ b/tests/Benchmark.hs
@@ -0,0 +1,1461 @@
+{-# OPTIONS -fbang-patterns #-}
+module Main (main) where
+
+import qualified Data.ByteString.Lazy as L
+import Data.Binary
+import Data.Binary.Put
+import Data.Binary.Get
+
+import Control.Exception
+import System.CPUTime
+import Numeric
+import Text.Printf
+import System.Environment
+
+import MemBench
+
+data Endian
+    = Big
+    | Little
+    | Host
+    deriving (Eq,Ord,Show)
+
+main :: IO ()
+main = do
+  mb <- getArgs >>= readIO . head
+  memBench (mb*10) 
+  putStrLn ""
+  putStrLn "Binary (de)serialisation benchmarks:"
+
+  -- do bytewise 
+  sequence_
+    [ test wordSize chunkSize Host mb
+    | wordSize  <- [1]
+    , chunkSize <- [16] --1,2,4,8,16]
+    ]
+
+  -- now Word16 .. Word64
+  sequence_
+    [ test wordSize chunkSize end mb
+    | wordSize  <- [2,4,8]
+    , chunkSize <- [16]
+    , end       <- [Host] -- ,Big,Little]
+    ]
+
+------------------------------------------------------------------------
+
+time :: IO a -> IO Double
+time action = do
+    start <- getCPUTime
+    action
+    end   <- getCPUTime
+    return $! (fromIntegral (end - start)) / (10^12)
+
+------------------------------------------------------------------------
+
+test :: Int -> Int -> Endian -> Int -> IO ()
+test wordSize chunkSize end mb = do
+    let bytes :: Int
+        bytes = mb * 2^20
+        iterations = bytes `div` wordSize
+        bs  = runPut (doPut wordSize chunkSize end iterations)
+        sum = runGet (doGet wordSize chunkSize end iterations) bs
+
+    case (chunkSize,end) of (1,Host) -> putStrLn "" ; _ -> return ()
+
+    printf "%dMB of Word%-2d in chunks of %2d (%6s endian): "
+        (mb :: Int) (8 * wordSize :: Int) (chunkSize :: Int) (show end)
+
+    putSeconds <- time $ evaluate (L.length bs)
+    getSeconds <- time $ evaluate sum
+--    print (L.length bs, sum)
+    let putThroughput = fromIntegral mb / putSeconds
+        getThroughput = fromIntegral mb / getSeconds
+
+    printf "%6.1f MB/s write, %6.1f MB/s read, %5.1f get/put-ratio\n"
+           putThroughput
+           getThroughput
+           (getThroughput/putThroughput)
+
+------------------------------------------------------------------------
+
+doPut :: Int -> Int -> Endian -> Int -> Put
+doPut wordSize chunkSize end = case (wordSize, chunkSize, end) of
+    (1, 1,_)   -> putWord8N1
+    (1, 2,_)   -> putWord8N2
+    (1, 4,_)   -> putWord8N4
+    (1, 8,_)   -> putWord8N8
+    (1, 16, _) -> putWord8N16
+
+    (2, 1,  Big)    -> putWord16N1Big
+    (2, 2,  Big)    -> putWord16N2Big
+    (2, 4,  Big)    -> putWord16N4Big
+    (2, 8,  Big)    -> putWord16N8Big
+    (2, 16, Big)    -> putWord16N16Big
+    (2, 1,  Little) -> putWord16N1Little
+    (2, 2,  Little) -> putWord16N2Little
+    (2, 4,  Little) -> putWord16N4Little
+    (2, 8,  Little) -> putWord16N8Little
+    (2, 16, Little) -> putWord16N16Little
+    (2, 1,  Host)   -> putWord16N1Host
+    (2, 2,  Host)   -> putWord16N2Host
+    (2, 4,  Host)   -> putWord16N4Host
+    (2, 8,  Host)   -> putWord16N8Host
+    (2, 16, Host)   -> putWord16N16Host
+
+    (4, 1,  Big)    -> putWord32N1Big
+    (4, 2,  Big)    -> putWord32N2Big
+    (4, 4,  Big)    -> putWord32N4Big
+    (4, 8,  Big)    -> putWord32N8Big
+    (4, 16, Big)    -> putWord32N16Big
+    (4, 1,  Little) -> putWord32N1Little
+    (4, 2,  Little) -> putWord32N2Little
+    (4, 4,  Little) -> putWord32N4Little
+    (4, 8,  Little) -> putWord32N8Little
+    (4, 16, Little) -> putWord32N16Little
+    (4, 1,  Host)   -> putWord32N1Host
+    (4, 2,  Host)   -> putWord32N2Host
+    (4, 4,  Host)   -> putWord32N4Host
+    (4, 8,  Host)   -> putWord32N8Host
+    (4, 16, Host)   -> putWord32N16Host
+
+    (8, 1,  Host)        -> putWord64N1Host
+    (8, 2,  Host)        -> putWord64N2Host
+    (8, 4,  Host)        -> putWord64N4Host
+    (8, 8,  Host)        -> putWord64N8Host
+    (8, 16, Host)        -> putWord64N16Host
+    (8, 1,  Big)         -> putWord64N1Big
+    (8, 2,  Big)         -> putWord64N2Big
+    (8, 4,  Big)         -> putWord64N4Big
+    (8, 8,  Big)         -> putWord64N8Big
+    (8, 16, Big)         -> putWord64N16Big
+    (8, 1,  Little)      -> putWord64N1Little
+    (8, 2,  Little)      -> putWord64N2Little
+    (8, 4,  Little)      -> putWord64N4Little
+    (8, 8,  Little)      -> putWord64N8Little
+    (8, 16, Little)      -> putWord64N16Little
+
+------------------------------------------------------------------------
+
+doGet :: Int -> Int -> Endian -> Int -> Get Int
+doGet wordSize chunkSize end =
+  case (wordSize, chunkSize, end) of
+    (1, 1,_)  -> fmap fromIntegral . getWord8N1
+    (1, 2,_)  -> fmap fromIntegral . getWord8N2
+    (1, 4,_)  -> fmap fromIntegral . getWord8N4
+    (1, 8,_)  -> fmap fromIntegral . getWord8N8
+    (1, 16,_) -> fmap fromIntegral . getWord8N16
+
+    (2, 1,Big)      -> fmap fromIntegral . getWord16N1Big
+    (2, 2,Big)      -> fmap fromIntegral . getWord16N2Big
+    (2, 4,Big)      -> fmap fromIntegral . getWord16N4Big
+    (2, 8,Big)      -> fmap fromIntegral . getWord16N8Big
+    (2, 16,Big)     -> fmap fromIntegral . getWord16N16Big
+    (2, 1,Little)   -> fmap fromIntegral . getWord16N1Little
+    (2, 2,Little)   -> fmap fromIntegral . getWord16N2Little
+    (2, 4,Little)   -> fmap fromIntegral . getWord16N4Little
+    (2, 8,Little)   -> fmap fromIntegral . getWord16N8Little
+    (2, 16,Little)  -> fmap fromIntegral . getWord16N16Little
+    (2, 1,Host)     -> fmap fromIntegral . getWord16N1Host
+    (2, 2,Host)     -> fmap fromIntegral . getWord16N2Host
+    (2, 4,Host)     -> fmap fromIntegral . getWord16N4Host
+    (2, 8,Host)     -> fmap fromIntegral . getWord16N8Host
+    (2, 16,Host)    -> fmap fromIntegral . getWord16N16Host
+
+    (4, 1,Big)      -> fmap fromIntegral . getWord32N1Big
+    (4, 2,Big)      -> fmap fromIntegral . getWord32N2Big
+    (4, 4,Big)      -> fmap fromIntegral . getWord32N4Big
+    (4, 8,Big)      -> fmap fromIntegral . getWord32N8Big
+    (4, 16,Big)     -> fmap fromIntegral . getWord32N16Big
+    (4, 1,Little)   -> fmap fromIntegral . getWord32N1Little
+    (4, 2,Little)   -> fmap fromIntegral . getWord32N2Little
+    (4, 4,Little)   -> fmap fromIntegral . getWord32N4Little
+    (4, 8,Little)   -> fmap fromIntegral . getWord32N8Little
+    (4, 16,Little)  -> fmap fromIntegral . getWord32N16Little
+    (4, 1,Host)     -> fmap fromIntegral . getWord32N1Host
+    (4, 2,Host)     -> fmap fromIntegral . getWord32N2Host
+    (4, 4,Host)     -> fmap fromIntegral . getWord32N4Host
+    (4, 8,Host)     -> fmap fromIntegral . getWord32N8Host
+    (4, 16,Host)    -> fmap fromIntegral . getWord32N16Host
+
+    (8, 1,Host)     -> fmap fromIntegral . getWord64N1Host
+    (8, 2,Host)     -> fmap fromIntegral . getWord64N2Host
+    (8, 4,Host)     -> fmap fromIntegral . getWord64N4Host
+    (8, 8,Host)     -> fmap fromIntegral . getWord64N8Host
+    (8, 16,Host)    -> fmap fromIntegral . getWord64N16Host
+    (8, 1,Big)      -> fmap fromIntegral . getWord64N1Big
+    (8, 2,Big)      -> fmap fromIntegral . getWord64N2Big
+    (8, 4,Big)      -> fmap fromIntegral . getWord64N4Big
+    (8, 8,Big)      -> fmap fromIntegral . getWord64N8Big
+    (8, 16,Big)     -> fmap fromIntegral . getWord64N16Big
+    (8, 1,Little)   -> fmap fromIntegral . getWord64N1Little
+    (8, 2,Little)   -> fmap fromIntegral . getWord64N2Little
+    (8, 4,Little)   -> fmap fromIntegral . getWord64N4Little
+    (8, 8,Little)   -> fmap fromIntegral . getWord64N8Little
+    (8, 16,Little)  -> fmap fromIntegral . getWord64N16Little
+
+------------------------------------------------------------------------
+
+putWord8N1 bytes = loop 0 0
+  where loop :: Word8 -> Int -> Put
+        loop !s !n | n == bytes = return ()
+                   | otherwise  = do putWord8 s
+                                     loop (s+1) (n+1)
+
+putWord8N2 = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord8 (s+0)
+          putWord8 (s+1)
+          loop (s+2) (n-2)
+
+putWord8N4 = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord8 (s+0)
+          putWord8 (s+1)
+          putWord8 (s+2)
+          putWord8 (s+3)
+          loop (s+4) (n-4)
+
+putWord8N8 = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord8 (s+0)
+          putWord8 (s+1)
+          putWord8 (s+2)
+          putWord8 (s+3)
+          putWord8 (s+4)
+          putWord8 (s+5)
+          putWord8 (s+6)
+          putWord8 (s+7)
+          loop (s+8) (n-8)
+
+putWord8N16 = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord8 (s+0)
+          putWord8 (s+1)
+          putWord8 (s+2)
+          putWord8 (s+3)
+          putWord8 (s+4)
+          putWord8 (s+5)
+          putWord8 (s+6)
+          putWord8 (s+7)
+          putWord8 (s+8)
+          putWord8 (s+9)
+          putWord8 (s+10)
+          putWord8 (s+11)
+          putWord8 (s+12)
+          putWord8 (s+13)
+          putWord8 (s+14)
+          putWord8 (s+15)
+          loop (s+16) (n-16)
+
+------------------------------------------------------------------------
+-- Big endian, word16 writes
+
+putWord16N1Big = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord16be (s+0)
+          loop (s+1) (n-1)
+
+putWord16N2Big = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord16be (s+0)
+          putWord16be (s+1)
+          loop (s+2) (n-2)
+
+putWord16N4Big = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord16be (s+0)
+          putWord16be (s+1)
+          putWord16be (s+2)
+          putWord16be (s+3)
+          loop (s+4) (n-4)
+
+putWord16N8Big = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord16be (s+0)
+          putWord16be (s+1)
+          putWord16be (s+2)
+          putWord16be (s+3)
+          putWord16be (s+4)
+          putWord16be (s+5)
+          putWord16be (s+6)
+          putWord16be (s+7)
+          loop (s+8) (n-8)
+
+putWord16N16Big = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord16be (s+0)
+          putWord16be (s+1)
+          putWord16be (s+2)
+          putWord16be (s+3)
+          putWord16be (s+4)
+          putWord16be (s+5)
+          putWord16be (s+6)
+          putWord16be (s+7)
+          putWord16be (s+8)
+          putWord16be (s+9)
+          putWord16be (s+10)
+          putWord16be (s+11)
+          putWord16be (s+12)
+          putWord16be (s+13)
+          putWord16be (s+14)
+          putWord16be (s+15)
+          loop (s+16) (n-16)
+
+------------------------------------------------------------------------
+-- Little endian, word16 writes
+
+putWord16N1Little = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord16le (s+0)
+          loop (s+1) (n-1)
+
+putWord16N2Little = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord16le (s+0)
+          putWord16le (s+1)
+          loop (s+2) (n-2)
+
+putWord16N4Little = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord16le (s+0)
+          putWord16le (s+1)
+          putWord16le (s+2)
+          putWord16le (s+3)
+          loop (s+4) (n-4)
+
+putWord16N8Little = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord16le (s+0)
+          putWord16le (s+1)
+          putWord16le (s+2)
+          putWord16le (s+3)
+          putWord16le (s+4)
+          putWord16le (s+5)
+          putWord16le (s+6)
+          putWord16le (s+7)
+          loop (s+8) (n-8)
+
+putWord16N16Little = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord16le (s+0)
+          putWord16le (s+1)
+          putWord16le (s+2)
+          putWord16le (s+3)
+          putWord16le (s+4)
+          putWord16le (s+5)
+          putWord16le (s+6)
+          putWord16le (s+7)
+          putWord16le (s+8)
+          putWord16le (s+9)
+          putWord16le (s+10)
+          putWord16le (s+11)
+          putWord16le (s+12)
+          putWord16le (s+13)
+          putWord16le (s+14)
+          putWord16le (s+15)
+          loop (s+16) (n-16)
+
+------------------------------------------------------------------------
+-- Host endian, unaligned, word16 writes
+
+putWord16N1Host = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord16host (s+0)
+          loop (s+1) (n-1)
+
+putWord16N2Host = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord16host (s+0)
+          putWord16host (s+1)
+          loop (s+2) (n-2)
+
+putWord16N4Host = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord16host (s+0)
+          putWord16host (s+1)
+          putWord16host (s+2)
+          putWord16host (s+3)
+          loop (s+4) (n-4)
+
+putWord16N8Host = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord16host (s+0)
+          putWord16host (s+1)
+          putWord16host (s+2)
+          putWord16host (s+3)
+          putWord16host (s+4)
+          putWord16host (s+5)
+          putWord16host (s+6)
+          putWord16host (s+7)
+          loop (s+8) (n-8)
+
+putWord16N16Host = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord16host (s+0)
+          putWord16host (s+1)
+          putWord16host (s+2)
+          putWord16host (s+3)
+          putWord16host (s+4)
+          putWord16host (s+5)
+          putWord16host (s+6)
+          putWord16host (s+7)
+          putWord16host (s+8)
+          putWord16host (s+9)
+          putWord16host (s+10)
+          putWord16host (s+11)
+          putWord16host (s+12)
+          putWord16host (s+13)
+          putWord16host (s+14)
+          putWord16host (s+15)
+          loop (s+16) (n-16)
+
+------------------------------------------------------------------------
+
+putWord32N1Big = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord32be (s+0)
+          loop (s+1) (n-1)
+
+putWord32N2Big = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord32be (s+0)
+          putWord32be (s+1)
+          loop (s+2) (n-2)
+
+putWord32N4Big = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord32be (s+0)
+          putWord32be (s+1)
+          putWord32be (s+2)
+          putWord32be (s+3)
+          loop (s+4) (n-4)
+
+putWord32N8Big = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord32be (s+0)
+          putWord32be (s+1)
+          putWord32be (s+2)
+          putWord32be (s+3)
+          putWord32be (s+4)
+          putWord32be (s+5)
+          putWord32be (s+6)
+          putWord32be (s+7)
+          loop (s+8) (n-8)
+
+putWord32N16Big = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord32be (s+0)
+          putWord32be (s+1)
+          putWord32be (s+2)
+          putWord32be (s+3)
+          putWord32be (s+4)
+          putWord32be (s+5)
+          putWord32be (s+6)
+          putWord32be (s+7)
+          putWord32be (s+8)
+          putWord32be (s+9)
+          putWord32be (s+10)
+          putWord32be (s+11)
+          putWord32be (s+12)
+          putWord32be (s+13)
+          putWord32be (s+14)
+          putWord32be (s+15)
+          loop (s+16) (n-16)
+
+------------------------------------------------------------------------
+
+putWord32N1Little = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord32le (s+0)
+          loop (s+1) (n-1)
+
+putWord32N2Little = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord32le (s+0)
+          putWord32le (s+1)
+          loop (s+2) (n-2)
+
+putWord32N4Little = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord32le (s+0)
+          putWord32le (s+1)
+          putWord32le (s+2)
+          putWord32le (s+3)
+          loop (s+4) (n-4)
+
+putWord32N8Little = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord32le (s+0)
+          putWord32le (s+1)
+          putWord32le (s+2)
+          putWord32le (s+3)
+          putWord32le (s+4)
+          putWord32le (s+5)
+          putWord32le (s+6)
+          putWord32le (s+7)
+          loop (s+8) (n-8)
+
+putWord32N16Little = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord32le (s+0)
+          putWord32le (s+1)
+          putWord32le (s+2)
+          putWord32le (s+3)
+          putWord32le (s+4)
+          putWord32le (s+5)
+          putWord32le (s+6)
+          putWord32le (s+7)
+          putWord32le (s+8)
+          putWord32le (s+9)
+          putWord32le (s+10)
+          putWord32le (s+11)
+          putWord32le (s+12)
+          putWord32le (s+13)
+          putWord32le (s+14)
+          putWord32le (s+15)
+          loop (s+16) (n-16)
+
+------------------------------------------------------------------------
+
+putWord32N1Host = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord32host (s+0)
+          loop (s+1) (n-1)
+
+putWord32N2Host = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord32host (s+0)
+          putWord32host (s+1)
+          loop (s+2) (n-2)
+
+putWord32N4Host = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord32host (s+0)
+          putWord32host (s+1)
+          putWord32host (s+2)
+          putWord32host (s+3)
+          loop (s+4) (n-4)
+
+putWord32N8Host = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord32host (s+0)
+          putWord32host (s+1)
+          putWord32host (s+2)
+          putWord32host (s+3)
+          putWord32host (s+4)
+          putWord32host (s+5)
+          putWord32host (s+6)
+          putWord32host (s+7)
+          loop (s+8) (n-8)
+
+putWord32N16Host = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord32host (s+0)
+          putWord32host (s+1)
+          putWord32host (s+2)
+          putWord32host (s+3)
+          putWord32host (s+4)
+          putWord32host (s+5)
+          putWord32host (s+6)
+          putWord32host (s+7)
+          putWord32host (s+8)
+          putWord32host (s+9)
+          putWord32host (s+10)
+          putWord32host (s+11)
+          putWord32host (s+12)
+          putWord32host (s+13)
+          putWord32host (s+14)
+          putWord32host (s+15)
+          loop (s+16) (n-16)
+
+------------------------------------------------------------------------
+
+putWord64N1Big = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord64be (s+0)
+          loop (s+1) (n-1)
+
+putWord64N2Big = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord64be (s+0)
+          putWord64be (s+1)
+          loop (s+2) (n-2)
+
+putWord64N4Big = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord64be (s+0)
+          putWord64be (s+1)
+          putWord64be (s+2)
+          putWord64be (s+3)
+          loop (s+4) (n-4)
+
+putWord64N8Big = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord64be (s+0)
+          putWord64be (s+1)
+          putWord64be (s+2)
+          putWord64be (s+3)
+          putWord64be (s+4)
+          putWord64be (s+5)
+          putWord64be (s+6)
+          putWord64be (s+7)
+          loop (s+8) (n-8)
+
+putWord64N16Big = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord64be (s+0)
+          putWord64be (s+1)
+          putWord64be (s+2)
+          putWord64be (s+3)
+          putWord64be (s+4)
+          putWord64be (s+5)
+          putWord64be (s+6)
+          putWord64be (s+7)
+          putWord64be (s+8)
+          putWord64be (s+9)
+          putWord64be (s+10)
+          putWord64be (s+11)
+          putWord64be (s+12)
+          putWord64be (s+13)
+          putWord64be (s+14)
+          putWord64be (s+15)
+          loop (s+16) (n-16)
+
+------------------------------------------------------------------------
+
+putWord64N1Little = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord64le (s+0)
+          loop (s+1) (n-1)
+
+putWord64N2Little = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord64le (s+0)
+          putWord64le (s+1)
+          loop (s+2) (n-2)
+
+putWord64N4Little = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord64le (s+0)
+          putWord64le (s+1)
+          putWord64le (s+2)
+          putWord64le (s+3)
+          loop (s+4) (n-4)
+
+putWord64N8Little = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord64le (s+0)
+          putWord64le (s+1)
+          putWord64le (s+2)
+          putWord64le (s+3)
+          putWord64le (s+4)
+          putWord64le (s+5)
+          putWord64le (s+6)
+          putWord64le (s+7)
+          loop (s+8) (n-8)
+
+putWord64N16Little = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord64le (s+0)
+          putWord64le (s+1)
+          putWord64le (s+2)
+          putWord64le (s+3)
+          putWord64le (s+4)
+          putWord64le (s+5)
+          putWord64le (s+6)
+          putWord64le (s+7)
+          putWord64le (s+8)
+          putWord64le (s+9)
+          putWord64le (s+10)
+          putWord64le (s+11)
+          putWord64le (s+12)
+          putWord64le (s+13)
+          putWord64le (s+14)
+          putWord64le (s+15)
+          loop (s+16) (n-16)
+
+------------------------------------------------------------------------
+
+putWord64N1Host = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord64host (s+0)
+          loop (s+1) (n-1)
+
+putWord64N2Host = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord64host (s+0)
+          putWord64host (s+1)
+          loop (s+2) (n-2)
+
+putWord64N4Host = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord64host (s+0)
+          putWord64host (s+1)
+          putWord64host (s+2)
+          putWord64host (s+3)
+          loop (s+4) (n-4)
+
+putWord64N8Host = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord64host (s+0)
+          putWord64host (s+1)
+          putWord64host (s+2)
+          putWord64host (s+3)
+          putWord64host (s+4)
+          putWord64host (s+5)
+          putWord64host (s+6)
+          putWord64host (s+7)
+          loop (s+8) (n-8)
+
+putWord64N16Host = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord64host (s+0)
+          putWord64host (s+1)
+          putWord64host (s+2)
+          putWord64host (s+3)
+          putWord64host (s+4)
+          putWord64host (s+5)
+          putWord64host (s+6)
+          putWord64host (s+7)
+          putWord64host (s+8)
+          putWord64host (s+9)
+          putWord64host (s+10)
+          putWord64host (s+11)
+          putWord64host (s+12)
+          putWord64host (s+13)
+          putWord64host (s+14)
+          putWord64host (s+15)
+          loop (s+16) (n-16)
+
+------------------------------------------------------------------------
+------------------------------------------------------------------------
+
+getWord8N1 = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord8
+          loop (s+s0) (n-1)
+
+getWord8N2 = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord8
+          s1 <- getWord8
+          loop (s+s0+s1) (n-2)
+
+getWord8N4 = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord8
+          s1 <- getWord8
+          s2 <- getWord8
+          s3 <- getWord8
+          loop (s+s0+s1+s2+s3) (n-4)
+
+getWord8N8 = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord8
+          s1 <- getWord8
+          s2 <- getWord8
+          s3 <- getWord8
+          s4 <- getWord8
+          s5 <- getWord8
+          s6 <- getWord8
+          s7 <- getWord8
+          loop (s+s0+s1+s2+s3+s4+s5+s6+s7) (n-8)
+
+getWord8N16 = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord8
+          s1 <- getWord8
+          s2 <- getWord8
+          s3 <- getWord8
+          s4 <- getWord8
+          s5 <- getWord8
+          s6 <- getWord8
+          s7 <- getWord8
+          s8 <- getWord8
+          s9 <- getWord8
+          s10 <- getWord8
+          s11 <- getWord8
+          s12 <- getWord8
+          s13 <- getWord8
+          s14 <- getWord8
+          s15 <- getWord8
+          loop (s+s0+s1+s2+s3+s4+s5+s6+s7+s9+s10+s11+s12+s13+s14+s15) (n-16)
+
+------------------------------------------------------------------------
+
+getWord16N1Big = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord16be
+          loop (s+s0) (n-1)
+
+getWord16N2Big = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord16be
+          s1 <- getWord16be
+          loop (s+s0+s1) (n-2)
+
+getWord16N4Big = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord16be
+          s1 <- getWord16be
+          s2 <- getWord16be
+          s3 <- getWord16be
+          loop (s+s0+s1+s2+s3) (n-4)
+
+getWord16N8Big = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord16be
+          s1 <- getWord16be
+          s2 <- getWord16be
+          s3 <- getWord16be
+          s4 <- getWord16be
+          s5 <- getWord16be
+          s6 <- getWord16be
+          s7 <- getWord16be
+          loop (s+s0+s1+s2+s3+s4+s5+s6+s7) (n-8)
+
+getWord16N16Big = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord16be
+          s1 <- getWord16be
+          s2 <- getWord16be
+          s3 <- getWord16be
+          s4 <- getWord16be
+          s5 <- getWord16be
+          s6 <- getWord16be
+          s7 <- getWord16be
+          s8 <- getWord16be
+          s9 <- getWord16be
+          s10 <- getWord16be
+          s11 <- getWord16be
+          s12 <- getWord16be
+          s13 <- getWord16be
+          s14 <- getWord16be
+          s15 <- getWord16be
+          loop (s+s0+s1+s2+s3+s4+s5+s6+s7+s9+s10+s11+s12+s13+s14+s15) (n-16)
+
+------------------------------------------------------------------------
+
+getWord16N1Little = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord16le
+          loop (s+s0) (n-1)
+
+getWord16N2Little = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord16le
+          s1 <- getWord16le
+          loop (s+s0+s1) (n-2)
+
+getWord16N4Little = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord16le
+          s1 <- getWord16le
+          s2 <- getWord16le
+          s3 <- getWord16le
+          loop (s+s0+s1+s2+s3) (n-4)
+
+getWord16N8Little = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord16le
+          s1 <- getWord16le
+          s2 <- getWord16le
+          s3 <- getWord16le
+          s4 <- getWord16le
+          s5 <- getWord16le
+          s6 <- getWord16le
+          s7 <- getWord16le
+          loop (s+s0+s1+s2+s3+s4+s5+s6+s7) (n-8)
+
+getWord16N16Little = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord16le
+          s1 <- getWord16le
+          s2 <- getWord16le
+          s3 <- getWord16le
+          s4 <- getWord16le
+          s5 <- getWord16le
+          s6 <- getWord16le
+          s7 <- getWord16le
+          s8 <- getWord16le
+          s9 <- getWord16le
+          s10 <- getWord16le
+          s11 <- getWord16le
+          s12 <- getWord16le
+          s13 <- getWord16le
+          s14 <- getWord16le
+          s15 <- getWord16le
+          loop (s+s0+s1+s2+s3+s4+s5+s6+s7+s9+s10+s11+s12+s13+s14+s15) (n-16)
+
+------------------------------------------------------------------------
+
+getWord16N1Host = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord16host
+          loop (s+s0) (n-1)
+
+getWord16N2Host = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord16host
+          s1 <- getWord16host
+          loop (s+s0+s1) (n-2)
+
+getWord16N4Host = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord16host
+          s1 <- getWord16host
+          s2 <- getWord16host
+          s3 <- getWord16host
+          loop (s+s0+s1+s2+s3) (n-4)
+
+getWord16N8Host = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord16host
+          s1 <- getWord16host
+          s2 <- getWord16host
+          s3 <- getWord16host
+          s4 <- getWord16host
+          s5 <- getWord16host
+          s6 <- getWord16host
+          s7 <- getWord16host
+          loop (s+s0+s1+s2+s3+s4+s5+s6+s7) (n-8)
+
+getWord16N16Host = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord16host
+          s1 <- getWord16host
+          s2 <- getWord16host
+          s3 <- getWord16host
+          s4 <- getWord16host
+          s5 <- getWord16host
+          s6 <- getWord16host
+          s7 <- getWord16host
+          s8 <- getWord16host
+          s9 <- getWord16host
+          s10 <- getWord16host
+          s11 <- getWord16host
+          s12 <- getWord16host
+          s13 <- getWord16host
+          s14 <- getWord16host
+          s15 <- getWord16host
+          loop (s+s0+s1+s2+s3+s4+s5+s6+s7+s9+s10+s11+s12+s13+s14+s15) (n-16)
+
+------------------------------------------------------------------------
+
+getWord32N1Big = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord32be
+          loop (s+s0) (n-1)
+
+getWord32N2Big = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord32be
+          s1 <- getWord32be
+          loop (s+s0+s1) (n-2)
+
+getWord32N4Big = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord32be
+          s1 <- getWord32be
+          s2 <- getWord32be
+          s3 <- getWord32be
+          loop (s+s0+s1+s2+s3) (n-4)
+
+getWord32N8Big = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord32be
+          s1 <- getWord32be
+          s2 <- getWord32be
+          s3 <- getWord32be
+          s4 <- getWord32be
+          s5 <- getWord32be
+          s6 <- getWord32be
+          s7 <- getWord32be
+          loop (s+s0+s1+s2+s3+s4+s5+s6+s7) (n-8)
+
+-- getWordhostN16 = loop 0
+getWord32N16Big = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord32be
+          s1 <- getWord32be
+          s2 <- getWord32be
+          s3 <- getWord32be
+          s4 <- getWord32be
+          s5 <- getWord32be
+          s6 <- getWord32be
+          s7 <- getWord32be
+          s8 <- getWord32be
+          s9 <- getWord32be
+          s10 <- getWord32be
+          s11 <- getWord32be
+          s12 <- getWord32be
+          s13 <- getWord32be
+          s14 <- getWord32be
+          s15 <- getWord32be
+          loop (s+s0+s1+s2+s3+s4+s5+s6+s7+s9+s10+s11+s12+s13+s14+s15) (n-16)
+
+------------------------------------------------------------------------
+
+getWord32N1Little = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord32le
+          loop (s+s0) (n-1)
+
+getWord32N2Little = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord32le
+          s1 <- getWord32le
+          loop (s+s0+s1) (n-2)
+
+getWord32N4Little = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord32le
+          s1 <- getWord32le
+          s2 <- getWord32le
+          s3 <- getWord32le
+          loop (s+s0+s1+s2+s3) (n-4)
+
+getWord32N8Little = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord32le
+          s1 <- getWord32le
+          s2 <- getWord32le
+          s3 <- getWord32le
+          s4 <- getWord32le
+          s5 <- getWord32le
+          s6 <- getWord32le
+          s7 <- getWord32le
+          loop (s+s0+s1+s2+s3+s4+s5+s6+s7) (n-8)
+
+-- getWordhostN16 = loop 0
+getWord32N16Little = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord32le
+          s1 <- getWord32le
+          s2 <- getWord32le
+          s3 <- getWord32le
+          s4 <- getWord32le
+          s5 <- getWord32le
+          s6 <- getWord32le
+          s7 <- getWord32le
+          s8 <- getWord32le
+          s9 <- getWord32le
+          s10 <- getWord32le
+          s11 <- getWord32le
+          s12 <- getWord32le
+          s13 <- getWord32le
+          s14 <- getWord32le
+          s15 <- getWord32le
+          loop (s+s0+s1+s2+s3+s4+s5+s6+s7+s9+s10+s11+s12+s13+s14+s15) (n-16)
+
+------------------------------------------------------------------------
+
+getWord32N1Host = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord32host
+          loop (s+s0) (n-1)
+
+getWord32N2Host = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord32host
+          s1 <- getWord32host
+          loop (s+s0+s1) (n-2)
+
+getWord32N4Host = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord32host
+          s1 <- getWord32host
+          s2 <- getWord32host
+          s3 <- getWord32host
+          loop (s+s0+s1+s2+s3) (n-4)
+
+getWord32N8Host = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord32host
+          s1 <- getWord32host
+          s2 <- getWord32host
+          s3 <- getWord32host
+          s4 <- getWord32host
+          s5 <- getWord32host
+          s6 <- getWord32host
+          s7 <- getWord32host
+          loop (s+s0+s1+s2+s3+s4+s5+s6+s7) (n-8)
+
+-- getWordhostN16 = loop 0
+getWord32N16Host = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord32host
+          s1 <- getWord32host
+          s2 <- getWord32host
+          s3 <- getWord32host
+          s4 <- getWord32host
+          s5 <- getWord32host
+          s6 <- getWord32host
+          s7 <- getWord32host
+          s8 <- getWord32host
+          s9 <- getWord32host
+          s10 <- getWord32host
+          s11 <- getWord32host
+          s12 <- getWord32host
+          s13 <- getWord32host
+          s14 <- getWord32host
+          s15 <- getWord32host
+          loop (s+s0+s1+s2+s3+s4+s5+s6+s7+s9+s10+s11+s12+s13+s14+s15) (n-16)
+
+------------------------------------------------------------------------
+
+getWord64N1Big = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord64be
+          loop (s+s0) (n-1)
+
+getWord64N2Big = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord64be
+          s1 <- getWord64be
+          loop (s+s0+s1) (n-2)
+
+getWord64N4Big = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord64be
+          s1 <- getWord64be
+          s2 <- getWord64be
+          s3 <- getWord64be
+          loop (s+s0+s1+s2+s3) (n-4)
+
+getWord64N8Big = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord64be
+          s1 <- getWord64be
+          s2 <- getWord64be
+          s3 <- getWord64be
+          s4 <- getWord64be
+          s5 <- getWord64be
+          s6 <- getWord64be
+          s7 <- getWord64be
+          loop (s+s0+s1+s2+s3+s4+s5+s6+s7) (n-8)
+
+getWord64N16Big = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord64be
+          s1 <- getWord64be
+          s2 <- getWord64be
+          s3 <- getWord64be
+          s4 <- getWord64be
+          s5 <- getWord64be
+          s6 <- getWord64be
+          s7 <- getWord64be
+          s8 <- getWord64be
+          s9 <- getWord64be
+          s10 <- getWord64be
+          s11 <- getWord64be
+          s12 <- getWord64be
+          s13 <- getWord64be
+          s14 <- getWord64be
+          s15 <- getWord64be
+          loop (s+s0+s1+s2+s3+s4+s5+s6+s7+s9+s10+s11+s12+s13+s14+s15) (n-16)
+
+------------------------------------------------------------------------
+
+getWord64N1Little = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord64le
+          loop (s+s0) (n-1)
+
+getWord64N2Little = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord64le
+          s1 <- getWord64le
+          loop (s+s0+s1) (n-2)
+
+getWord64N4Little = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord64le
+          s1 <- getWord64le
+          s2 <- getWord64le
+          s3 <- getWord64le
+          loop (s+s0+s1+s2+s3) (n-4)
+
+getWord64N8Little = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord64le
+          s1 <- getWord64le
+          s2 <- getWord64le
+          s3 <- getWord64le
+          s4 <- getWord64le
+          s5 <- getWord64le
+          s6 <- getWord64le
+          s7 <- getWord64le
+          loop (s+s0+s1+s2+s3+s4+s5+s6+s7) (n-8)
+
+getWord64N16Little = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord64le
+          s1 <- getWord64le
+          s2 <- getWord64le
+          s3 <- getWord64le
+          s4 <- getWord64le
+          s5 <- getWord64le
+          s6 <- getWord64le
+          s7 <- getWord64le
+          s8 <- getWord64le
+          s9 <- getWord64le
+          s10 <- getWord64le
+          s11 <- getWord64le
+          s12 <- getWord64le
+          s13 <- getWord64le
+          s14 <- getWord64le
+          s15 <- getWord64le
+          loop (s+s0+s1+s2+s3+s4+s5+s6+s7+s9+s10+s11+s12+s13+s14+s15) (n-16)
+
+------------------------------------------------------------------------
+
+getWord64N1Host = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord64host
+          loop (s+s0) (n-1)
+
+getWord64N2Host = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord64host
+          s1 <- getWord64host
+          loop (s+s0+s1) (n-2)
+
+getWord64N4Host = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord64host
+          s1 <- getWord64host
+          s2 <- getWord64host
+          s3 <- getWord64host
+          loop (s+s0+s1+s2+s3) (n-4)
+
+getWord64N8Host = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord64host
+          s1 <- getWord64host
+          s2 <- getWord64host
+          s3 <- getWord64host
+          s4 <- getWord64host
+          s5 <- getWord64host
+          s6 <- getWord64host
+          s7 <- getWord64host
+          loop (s+s0+s1+s2+s3+s4+s5+s6+s7) (n-8)
+
+getWord64N16Host = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord64host
+          s1 <- getWord64host
+          s2 <- getWord64host
+          s3 <- getWord64host
+          s4 <- getWord64host
+          s5 <- getWord64host
+          s6 <- getWord64host
+          s7 <- getWord64host
+          s8 <- getWord64host
+          s9 <- getWord64host
+          s10 <- getWord64host
+          s11 <- getWord64host
+          s12 <- getWord64host
+          s13 <- getWord64host
+          s14 <- getWord64host
+          s15 <- getWord64host
+          loop (s+s0+s1+s2+s3+s4+s5+s6+s7+s9+s10+s11+s12+s13+s14+s15) (n-16)
diff --git a/tests/CBenchmark.c b/tests/CBenchmark.c
new file mode 100644
--- /dev/null
+++ b/tests/CBenchmark.c
@@ -0,0 +1,39 @@
+#include "CBenchmark.h"
+
+void bytewrite(unsigned char *a, int bytes) {
+  unsigned char n = 0;
+  int i = 0;
+  int iterations = bytes;
+  while (i < iterations) {
+    a[i++] = n++;
+  }
+}
+
+unsigned char byteread(unsigned char *a, int bytes) {
+  unsigned char n = 0;
+  int i = 0;
+  int iterations = bytes;
+  while (i < iterations) {
+    n += a[i++];
+  }
+  return n;
+}
+
+void wordwrite(unsigned long *a, int bytes) {
+  unsigned long n = 0;
+  int i = 0;
+  int iterations = bytes / sizeof(unsigned long) ;
+  while (i < iterations) {
+    a[i++] = n++;
+  }
+}
+
+unsigned int wordread(unsigned long *a, int bytes) {
+  unsigned long n = 0;
+  int i = 0;
+  int iterations = bytes / sizeof(unsigned long);
+  while (i < iterations) {
+    n += a[i++];
+  }
+  return n;
+}
diff --git a/tests/CBenchmark.h b/tests/CBenchmark.h
new file mode 100644
--- /dev/null
+++ b/tests/CBenchmark.h
@@ -0,0 +1,4 @@
+void bytewrite(unsigned char *a, int bytes);
+unsigned char byteread(unsigned char *a, int bytes);
+void wordwrite(unsigned long *a, int bytes);
+unsigned int wordread(unsigned long *a, int bytes);
diff --git a/tests/HeapUse.hs b/tests/HeapUse.hs
new file mode 100644
--- /dev/null
+++ b/tests/HeapUse.hs
@@ -0,0 +1,17 @@
+-- Checks heap behavior of getBytes
+
+import Data.Binary.Get (runGet, getBytes)
+
+import Control.Monad (liftM)
+import qualified Data.ByteString.Lazy as L
+
+main = do
+       let x = (L.take 110000042 $ L.iterate (+1) 0)
+       mapM_ (print . L.length) (chunks 20000000 x)
+
+chunks n = runGet (unfoldM f)
+  where f = do x <- getBytes 20000000 
+               return $ if L.null x then Nothing else Just x
+
+unfoldM :: Monad m => m (Maybe a) -> m [a]
+unfoldM f = f >>= maybe (return []) (\x -> liftM (x:) (unfoldM f))
diff --git a/tests/Makefile b/tests/Makefile
new file mode 100644
--- /dev/null
+++ b/tests/Makefile
@@ -0,0 +1,34 @@
+all: compiled
+     
+interpreted:
+	runhaskell QC.hs 1000
+
+compiled:
+	ghc --make -fhpc -O QC.hs -o qc -no-recomp -threaded
+	./qc 500 +RTS -qw -N2
+
+bench:: Benchmark.hs MemBench.hs CBenchmark.o
+	ghc --make -O2 Benchmark.hs -fasm CBenchmark.o -o bench -no-recomp
+	./bench 100
+
+bench-nb::
+	ghc --make -O2 -fliberate-case-threshold=1000 NewBenchmark.hs -fasm -o bench-nb
+	./bench-nb 
+
+CBenchmark.o: CBenchmark.c
+	gcc -O -c $< -o $@
+
+hugs:
+	runhugs -98 QC.hs  
+
+
+HeapUse: HeapUse.hs
+	ghc --make -O $^ -fasm -o $@
+
+heap: HeapUse
+	./HeapUse +RTS -M10M -t/dev/stderr -RTS
+
+clean:
+	rm -f *.o *.hi qc bench bench-nb *~
+
+.PHONY: clean bench bench-nb
diff --git a/tests/MemBench.hs b/tests/MemBench.hs
new file mode 100644
--- /dev/null
+++ b/tests/MemBench.hs
@@ -0,0 +1,85 @@
+{-# OPTIONS_GHC -fffi -fbang-patterns #-}
+module MemBench (memBench) where
+
+import Foreign
+import Foreign.C
+
+import Control.Exception
+import System.CPUTime
+import Numeric
+
+memBench :: Int -> IO ()
+memBench mb = do
+  let bytes = mb * 2^20
+  allocaBytes bytes $ \ptr -> do
+    let bench label test = do
+          seconds <- time $ test (castPtr ptr) (fromIntegral bytes)
+          let throughput = fromIntegral mb / seconds
+          putStrLn $ show mb ++ "MB of " ++ label
+                  ++ " in " ++ showFFloat (Just 3) seconds "s, at: "
+                  ++ showFFloat (Just 1) throughput "MB/s"
+    bench "setup        " c_wordwrite
+    putStrLn ""
+    putStrLn "C memory throughput benchmarks:"
+    bench "bytes written" c_bytewrite
+    bench "bytes read   " c_byteread
+    bench "words written" c_wordwrite
+    bench "words read   " c_wordread
+    putStrLn ""
+    putStrLn "Haskell memory throughput benchmarks:"
+    bench "bytes written" hs_bytewrite
+    bench "bytes read   " hs_byteread
+    bench "words written" hs_wordwrite
+    bench "words read   " hs_wordread
+
+hs_bytewrite  :: Ptr CUChar -> Int -> IO ()
+hs_bytewrite !ptr bytes = loop 0 0
+  where iterations = bytes
+        loop :: Int -> CUChar -> IO ()
+        loop !i !n | i == iterations = return ()
+                   | otherwise = do pokeByteOff ptr i n
+                                    loop (i+1) (n+1)
+
+hs_byteread  :: Ptr CUChar -> Int -> IO CUChar
+hs_byteread !ptr bytes = loop 0 0
+  where iterations = bytes
+        loop :: Int -> CUChar -> IO CUChar
+        loop !i !n | i == iterations = return n
+                   | otherwise = do x <- peekByteOff ptr i
+                                    loop (i+1) (n+x)
+
+hs_wordwrite :: Ptr CULong -> Int -> IO ()
+hs_wordwrite !ptr bytes = loop 0 0
+  where iterations = bytes `div` sizeOf (undefined :: CULong)
+        loop :: Int -> CULong -> IO ()
+        loop !i !n | i == iterations = return ()
+                   | otherwise = do pokeByteOff ptr i n
+                                    loop (i+1) (n+1)
+
+hs_wordread  :: Ptr CULong -> Int -> IO CULong
+hs_wordread !ptr bytes = loop 0 0
+  where iterations = bytes `div` sizeOf (undefined :: CULong)
+        loop :: Int -> CULong -> IO CULong
+        loop !i !n | i == iterations = return n
+                   | otherwise = do x <- peekByteOff ptr i
+                                    loop (i+1) (n+x)
+
+
+foreign import ccall unsafe "CBenchmark.h byteread"
+  c_byteread :: Ptr CUChar -> CInt -> IO ()
+
+foreign import ccall unsafe "CBenchmark.h bytewrite"
+  c_bytewrite :: Ptr CUChar -> CInt -> IO ()
+
+foreign import ccall unsafe "CBenchmark.h wordread"
+  c_wordread :: Ptr CUInt -> CInt -> IO ()
+
+foreign import ccall unsafe "CBenchmark.h wordwrite"
+  c_wordwrite :: Ptr CUInt -> CInt -> IO ()
+
+time :: IO a -> IO Double
+time action = do
+    start <- getCPUTime
+    action
+    end   <- getCPUTime
+    return $! (fromIntegral (end - start)) / (10^12)
diff --git a/tests/NewBenchmark.hs b/tests/NewBenchmark.hs
new file mode 100644
--- /dev/null
+++ b/tests/NewBenchmark.hs
@@ -0,0 +1,625 @@
+--
+-- benchmark NewBinary
+--
+
+module Main where
+
+import System.IO
+import Data.Word
+import NewBinary
+
+import Control.Exception
+import System.CPUTime
+import Numeric
+
+mb :: Int
+mb = 10
+
+main :: IO ()
+main = sequence_ 
+  [ test wordSize chunkSize mb
+  | wordSize  <- [1,2,4,8]
+  , chunkSize <- [1,2,4,8,16] ]
+
+time :: IO a -> IO Double
+time action = do
+    start <- getCPUTime
+    action
+    end   <- getCPUTime
+    return $! (fromIntegral (end - start)) / (10^12)
+
+test :: Int -> Int -> Int -> IO ()
+test wordSize chunkSize mb = do
+    let bytes :: Int
+        bytes = mb * 2^20
+        iterations = bytes `div` wordSize
+    putStr $ show mb ++ "MB of Word" ++ show (8 * wordSize)
+          ++ " in chunks of " ++ show chunkSize ++ ": "
+    h <- openBinMem bytes undefined
+    start <- tellBin h
+    putSeconds <- time $ do
+      doPut wordSize chunkSize h iterations
+--      BinPtr n _ <- tellBin h
+--      print n
+    getSeconds <- time $ do
+      seekBin h start
+      sum <- doGet wordSize chunkSize h iterations
+      evaluate sum
+--      BinPtr n _ <- tellBin h
+--      print (n, sum)
+    let putThroughput = fromIntegral mb / putSeconds
+        getThroughput = fromIntegral mb / getSeconds
+    putStrLn $ showFFloat (Just 2) putThroughput "MB/s write, "
+            ++ showFFloat (Just 2) getThroughput "MB/s read"
+
+doPut :: Int -> Int -> BinHandle -> Int -> IO ()
+doPut wordSize chunkSize =
+  case (wordSize, chunkSize) of
+    (1, 1)  -> putWord8N1
+    (1, 2)  -> putWord8N2
+    (1, 4)  -> putWord8N4
+    (1, 8)  -> putWord8N8
+    (1, 16) -> putWord8N16
+    (2, 1)  -> putWord16N1
+    (2, 2)  -> putWord16N2
+    (2, 4)  -> putWord16N4
+    (2, 8)  -> putWord16N8
+    (2, 16) -> putWord16N16
+    (4, 1)  -> putWord32N1
+    (4, 2)  -> putWord32N2
+    (4, 4)  -> putWord32N4
+    (4, 8)  -> putWord32N8
+    (4, 16) -> putWord32N16
+    (8, 1)  -> putWord64N1
+    (8, 2)  -> putWord64N2
+    (8, 4)  -> putWord64N4
+    (8, 8)  -> putWord64N8
+    (8, 16) -> putWord64N16
+
+putWord8 :: BinHandle -> Word8 -> IO ()
+putWord8 = put_
+{-# INLINE putWord8 #-}
+
+putWord16be :: BinHandle -> Word16 -> IO ()
+putWord16be = put_
+{-# INLINE putWord16be #-}
+
+putWord32be :: BinHandle -> Word32 -> IO ()
+putWord32be = put_
+{-# INLINE putWord32be #-}
+
+putWord64be :: BinHandle -> Word64 -> IO ()
+putWord64be = put_
+{-# INLINE putWord64be #-}
+
+getWord8 :: BinHandle -> IO Word8
+getWord8 = get
+{-# INLINE getWord8 #-}
+
+getWord16be :: BinHandle -> IO Word16
+getWord16be = get
+{-# INLINE getWord16be #-}
+
+getWord32be :: BinHandle -> IO Word32
+getWord32be = get
+{-# INLINE getWord32be #-}
+
+getWord64be :: BinHandle -> IO Word64
+getWord64be = get
+{-# INLINE getWord64be #-}
+
+putWord8N1 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord8 hnd (s+0)
+          loop (s+1) (n-1)
+
+putWord8N2 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord8 hnd (s+0)
+          putWord8 hnd (s+1)
+          loop (s+2) (n-2)
+
+putWord8N4 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord8 hnd (s+0)
+          putWord8 hnd (s+1)
+          putWord8 hnd (s+2)
+          putWord8 hnd (s+3)
+          loop (s+4) (n-4)
+
+putWord8N8 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord8 hnd (s+0)
+          putWord8 hnd (s+1)
+          putWord8 hnd (s+2)
+          putWord8 hnd (s+3)
+          putWord8 hnd (s+4)
+          putWord8 hnd (s+5)
+          putWord8 hnd (s+6)
+          putWord8 hnd (s+7)
+          loop (s+8) (n-8)
+
+putWord8N16 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord8 hnd (s+0)
+          putWord8 hnd (s+1)
+          putWord8 hnd (s+2)
+          putWord8 hnd (s+3)
+          putWord8 hnd (s+4)
+          putWord8 hnd (s+5)
+          putWord8 hnd (s+6)
+          putWord8 hnd (s+7)
+          putWord8 hnd (s+8)
+          putWord8 hnd (s+9)
+          putWord8 hnd (s+10)
+          putWord8 hnd (s+11)
+          putWord8 hnd (s+12)
+          putWord8 hnd (s+13)
+          putWord8 hnd (s+14)
+          putWord8 hnd (s+15)
+          loop (s+16) (n-16)
+
+
+putWord16N1 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord16be hnd (s+0)
+          loop (s+1) (n-1)
+
+putWord16N2 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord16be hnd (s+0)
+          putWord16be hnd (s+1)
+          loop (s+2) (n-2)
+
+putWord16N4 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord16be hnd (s+0)
+          putWord16be hnd (s+1)
+          putWord16be hnd (s+2)
+          putWord16be hnd (s+3)
+          loop (s+4) (n-4)
+
+putWord16N8 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord16be hnd (s+0)
+          putWord16be hnd (s+1)
+          putWord16be hnd (s+2)
+          putWord16be hnd (s+3)
+          putWord16be hnd (s+4)
+          putWord16be hnd (s+5)
+          putWord16be hnd (s+6)
+          putWord16be hnd (s+7)
+          loop (s+8) (n-8)
+
+putWord16N16 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord16be hnd (s+0)
+          putWord16be hnd (s+1)
+          putWord16be hnd (s+2)
+          putWord16be hnd (s+3)
+          putWord16be hnd (s+4)
+          putWord16be hnd (s+5)
+          putWord16be hnd (s+6)
+          putWord16be hnd (s+7)
+          putWord16be hnd (s+8)
+          putWord16be hnd (s+9)
+          putWord16be hnd (s+10)
+          putWord16be hnd (s+11)
+          putWord16be hnd (s+12)
+          putWord16be hnd (s+13)
+          putWord16be hnd (s+14)
+          putWord16be hnd (s+15)
+          loop (s+16) (n-16)
+
+
+putWord32N1 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord32be hnd (s+0)
+          loop (s+1) (n-1)
+
+putWord32N2 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord32be hnd (s+0)
+          putWord32be hnd (s+1)
+          loop (s+2) (n-2)
+
+putWord32N4 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord32be hnd (s+0)
+          putWord32be hnd (s+1)
+          putWord32be hnd (s+2)
+          putWord32be hnd (s+3)
+          loop (s+4) (n-4)
+
+putWord32N8 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord32be hnd (s+0)
+          putWord32be hnd (s+1)
+          putWord32be hnd (s+2)
+          putWord32be hnd (s+3)
+          putWord32be hnd (s+4)
+          putWord32be hnd (s+5)
+          putWord32be hnd (s+6)
+          putWord32be hnd (s+7)
+          loop (s+8) (n-8)
+
+putWord32N16 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord32be hnd (s+0)
+          putWord32be hnd (s+1)
+          putWord32be hnd (s+2)
+          putWord32be hnd (s+3)
+          putWord32be hnd (s+4)
+          putWord32be hnd (s+5)
+          putWord32be hnd (s+6)
+          putWord32be hnd (s+7)
+          putWord32be hnd (s+8)
+          putWord32be hnd (s+9)
+          putWord32be hnd (s+10)
+          putWord32be hnd (s+11)
+          putWord32be hnd (s+12)
+          putWord32be hnd (s+13)
+          putWord32be hnd (s+14)
+          putWord32be hnd (s+15)
+          loop (s+16) (n-16)
+
+putWord64N1 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord64be hnd (s+0)
+          loop (s+1) (n-1)
+
+putWord64N2 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord64be hnd (s+0)
+          putWord64be hnd (s+1)
+          loop (s+2) (n-2)
+
+putWord64N4 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord64be hnd (s+0)
+          putWord64be hnd (s+1)
+          putWord64be hnd (s+2)
+          putWord64be hnd (s+3)
+          loop (s+4) (n-4)
+
+putWord64N8 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord64be hnd (s+0)
+          putWord64be hnd (s+1)
+          putWord64be hnd (s+2)
+          putWord64be hnd (s+3)
+          putWord64be hnd (s+4)
+          putWord64be hnd (s+5)
+          putWord64be hnd (s+6)
+          putWord64be hnd (s+7)
+          loop (s+8) (n-8)
+
+putWord64N16 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop _ 0 = return ()
+        loop s n = do
+          putWord64be hnd (s+0)
+          putWord64be hnd (s+1)
+          putWord64be hnd (s+2)
+          putWord64be hnd (s+3)
+          putWord64be hnd (s+4)
+          putWord64be hnd (s+5)
+          putWord64be hnd (s+6)
+          putWord64be hnd (s+7)
+          putWord64be hnd (s+8)
+          putWord64be hnd (s+9)
+          putWord64be hnd (s+10)
+          putWord64be hnd (s+11)
+          putWord64be hnd (s+12)
+          putWord64be hnd (s+13)
+          putWord64be hnd (s+14)
+          putWord64be hnd (s+15)
+          loop (s+16) (n-16)
+
+doGet :: Int -> Int -> BinHandle -> Int ->  IO Int
+doGet wordSize chunkSize hnd =
+  case (wordSize, chunkSize) of
+    (1, 1)  -> fmap fromIntegral . getWord8N1 hnd
+    (1, 2)  -> fmap fromIntegral . getWord8N2 hnd
+    (1, 4)  -> fmap fromIntegral . getWord8N4 hnd
+    (1, 8)  -> fmap fromIntegral . getWord8N8 hnd
+    (1, 16) -> fmap fromIntegral . getWord8N16 hnd
+    (2, 1)  -> fmap fromIntegral . getWord16N1 hnd
+    (2, 2)  -> fmap fromIntegral . getWord16N2 hnd
+    (2, 4)  -> fmap fromIntegral . getWord16N4 hnd
+    (2, 8)  -> fmap fromIntegral . getWord16N8 hnd
+    (2, 16) -> fmap fromIntegral . getWord16N16 hnd
+    (4, 1)  -> fmap fromIntegral . getWord32N1 hnd
+    (4, 2)  -> fmap fromIntegral . getWord32N2 hnd
+    (4, 4)  -> fmap fromIntegral . getWord32N4 hnd
+    (4, 8)  -> fmap fromIntegral . getWord32N8 hnd
+    (4, 16) -> fmap fromIntegral . getWord32N16 hnd
+    (8, 1)  -> fmap fromIntegral . getWord64N1 hnd
+    (8, 2)  -> fmap fromIntegral . getWord64N2 hnd
+    (8, 4)  -> fmap fromIntegral . getWord64N4 hnd
+    (8, 8)  -> fmap fromIntegral . getWord64N8 hnd
+    (8, 16) -> fmap fromIntegral . getWord64N16 hnd
+
+getWord8N1 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord8 hnd
+          loop (s+s0) (n-1)
+
+getWord8N2 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord8 hnd
+          s1 <- getWord8 hnd
+          loop (s+s0+s1) (n-2)
+
+getWord8N4 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord8 hnd
+          s1 <- getWord8 hnd
+          s2 <- getWord8 hnd
+          s3 <- getWord8 hnd
+          loop (s+s0+s1+s2+s3) (n-4)
+
+getWord8N8 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord8 hnd
+          s1 <- getWord8 hnd
+          s2 <- getWord8 hnd
+          s3 <- getWord8 hnd
+          s4 <- getWord8 hnd
+          s5 <- getWord8 hnd
+          s6 <- getWord8 hnd
+          s7 <- getWord8 hnd
+          loop (s+s0+s1+s2+s3+s4+s5+s6+s7) (n-8)
+
+getWord8N16 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord8 hnd
+          s1 <- getWord8 hnd
+          s2 <- getWord8 hnd
+          s3 <- getWord8 hnd
+          s4 <- getWord8 hnd
+          s5 <- getWord8 hnd
+          s6 <- getWord8 hnd
+          s7 <- getWord8 hnd
+          s8 <- getWord8 hnd
+          s9 <- getWord8 hnd
+          s10 <- getWord8 hnd
+          s11 <- getWord8 hnd
+          s12 <- getWord8 hnd
+          s13 <- getWord8 hnd
+          s14 <- getWord8 hnd
+          s15 <- getWord8 hnd
+          loop (s+s0+s1+s2+s3+s4+s5+s6+s7+s9+s10+s11+s12+s13+s14+s15) (n-16)
+
+
+getWord16N1 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord16be hnd
+          loop (s+s0) (n-1)
+
+getWord16N2 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord16be hnd
+          s1 <- getWord16be hnd
+          loop (s+s0+s1) (n-2)
+
+getWord16N4 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord16be hnd
+          s1 <- getWord16be hnd
+          s2 <- getWord16be hnd
+          s3 <- getWord16be hnd
+          loop (s+s0+s1+s2+s3) (n-4)
+
+getWord16N8 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord16be hnd
+          s1 <- getWord16be hnd
+          s2 <- getWord16be hnd
+          s3 <- getWord16be hnd
+          s4 <- getWord16be hnd
+          s5 <- getWord16be hnd
+          s6 <- getWord16be hnd
+          s7 <- getWord16be hnd
+          loop (s+s0+s1+s2+s3+s4+s5+s6+s7) (n-8)
+
+getWord16N16 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord16be hnd
+          s1 <- getWord16be hnd
+          s2 <- getWord16be hnd
+          s3 <- getWord16be hnd
+          s4 <- getWord16be hnd
+          s5 <- getWord16be hnd
+          s6 <- getWord16be hnd
+          s7 <- getWord16be hnd
+          s8 <- getWord16be hnd
+          s9 <- getWord16be hnd
+          s10 <- getWord16be hnd
+          s11 <- getWord16be hnd
+          s12 <- getWord16be hnd
+          s13 <- getWord16be hnd
+          s14 <- getWord16be hnd
+          s15 <- getWord16be hnd
+          loop (s+s0+s1+s2+s3+s4+s5+s6+s7+s9+s10+s11+s12+s13+s14+s15) (n-16)
+
+
+getWord32N1 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord32be hnd
+          loop (s+s0) (n-1)
+
+getWord32N2 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord32be hnd
+          s1 <- getWord32be hnd
+          loop (s+s0+s1) (n-2)
+
+getWord32N4 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord32be hnd
+          s1 <- getWord32be hnd
+          s2 <- getWord32be hnd
+          s3 <- getWord32be hnd
+          loop (s+s0+s1+s2+s3) (n-4)
+
+getWord32N8 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord32be hnd
+          s1 <- getWord32be hnd
+          s2 <- getWord32be hnd
+          s3 <- getWord32be hnd
+          s4 <- getWord32be hnd
+          s5 <- getWord32be hnd
+          s6 <- getWord32be hnd
+          s7 <- getWord32be hnd
+          loop (s+s0+s1+s2+s3+s4+s5+s6+s7) (n-8)
+
+getWord32N16 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord32be hnd
+          s1 <- getWord32be hnd
+          s2 <- getWord32be hnd
+          s3 <- getWord32be hnd
+          s4 <- getWord32be hnd
+          s5 <- getWord32be hnd
+          s6 <- getWord32be hnd
+          s7 <- getWord32be hnd
+          s8 <- getWord32be hnd
+          s9 <- getWord32be hnd
+          s10 <- getWord32be hnd
+          s11 <- getWord32be hnd
+          s12 <- getWord32be hnd
+          s13 <- getWord32be hnd
+          s14 <- getWord32be hnd
+          s15 <- getWord32be hnd
+          loop (s+s0+s1+s2+s3+s4+s5+s6+s7+s9+s10+s11+s12+s13+s14+s15) (n-16)
+
+getWord64N1 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord64be hnd
+          loop (s+s0) (n-1)
+
+getWord64N2 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord64be hnd
+          s1 <- getWord64be hnd
+          loop (s+s0+s1) (n-2)
+
+getWord64N4 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord64be hnd
+          s1 <- getWord64be hnd
+          s2 <- getWord64be hnd
+          s3 <- getWord64be hnd
+          loop (s+s0+s1+s2+s3) (n-4)
+
+getWord64N8 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord64be hnd
+          s1 <- getWord64be hnd
+          s2 <- getWord64be hnd
+          s3 <- getWord64be hnd
+          s4 <- getWord64be hnd
+          s5 <- getWord64be hnd
+          s6 <- getWord64be hnd
+          s7 <- getWord64be hnd
+          loop (s+s0+s1+s2+s3+s4+s5+s6+s7) (n-8)
+
+getWord64N16 hnd = loop 0
+  where loop s n | s `seq` n `seq` False = undefined
+        loop s 0 = return s
+        loop s n = do
+          s0 <- getWord64be hnd
+          s1 <- getWord64be hnd
+          s2 <- getWord64be hnd
+          s3 <- getWord64be hnd
+          s4 <- getWord64be hnd
+          s5 <- getWord64be hnd
+          s6 <- getWord64be hnd
+          s7 <- getWord64be hnd
+          s8 <- getWord64be hnd
+          s9 <- getWord64be hnd
+          s10 <- getWord64be hnd
+          s11 <- getWord64be hnd
+          s12 <- getWord64be hnd
+          s13 <- getWord64be hnd
+          s14 <- getWord64be hnd
+          s15 <- getWord64be hnd
+          loop (s+s0+s1+s2+s3+s4+s5+s6+s7+s9+s10+s11+s12+s13+s14+s15) (n-16)
diff --git a/tests/NewBinary.hs b/tests/NewBinary.hs
new file mode 100644
--- /dev/null
+++ b/tests/NewBinary.hs
@@ -0,0 +1,1006 @@
+{-# OPTIONS -cpp -fglasgow-exts  #-}
+--
+-- (c) The University of Glasgow 2002
+--
+-- Binary I/O library, with special tweaks for GHC
+--
+-- Based on the nhc98 Binary library, which is copyright
+-- (c) Malcolm Wallace and Colin Runciman, University of York, 1998.
+-- Under the terms of the license for that software, we must tell you
+-- where you can obtain the original version of the Binary library, namely
+--     http://www.cs.york.ac.uk/fp/nhc98/
+
+module NewBinary
+  ( {-type-}  Bin,
+    {-class-} Binary(..),
+    {-type-}  BinHandle(..),
+
+   openBinIO, 
+   openBinIO_,
+   openBinMem,
+--   closeBin,
+
+--   getUserData,
+
+   seekBin,
+   tellBin,
+   tellBinByte,
+   castBin,
+
+   writeBinMem,
+   readBinMem,
+
+   isEOFBin,
+
+   -- for writing instances:
+   putByte,
+   getByte,
+
+   -- bit stuff
+   putBits,
+   getBits,
+   flushByte,
+   finishByte,
+   putMaybeInt,
+   getMaybeInt,
+
+   -- lazy Bin I/O
+   lazyGet,
+   lazyPut,
+
+   -- GHC only:
+   ByteArray(..),
+   getByteArray,
+   putByteArray,
+
+--   getBinFileWithDict,    -- :: Binary a => FilePath -> IO a
+--   putBinFileWithDict,    -- :: Binary a => FilePath -> Module -> a -> IO ()
+
+  ) where
+
+#include "MachDeps.h"
+
+import GHC.Exts
+import GHC.IOBase
+import GHC.Real
+import Data.Array.IO        ( IOUArray )
+import Data.Bits
+import Data.Int
+import Data.Word
+import Data.Char
+import Control.Monad
+import Control.Exception
+import Data.Array
+import Data.Array.IO
+import Data.Array.Base
+import System.IO as IO
+import System.IO.Error      ( mkIOError, eofErrorType )
+import GHC.Handle       
+import System.IO
+
+import GHC.Exts
+#if __GLASGOW_HASKELL__ >= 504
+import GHC.IOBase
+import Data.Word
+import Data.Bits
+#else
+import PrelIOBase
+import Word
+import Bits
+#endif
+
+#ifndef SIZEOF_HSINT
+#define SIZEOF_HSINT  INT_SIZE_IN_BYTES
+#endif
+
+#if __GLASGOW_HASKELL__ < 503
+type BinArray = MutableByteArray RealWorld Int
+newArray_ bounds     = stToIO (newCharArray bounds)
+unsafeWrite arr ix e = stToIO (writeWord8Array arr ix e)
+unsafeRead  arr ix   = stToIO (readWord8Array arr ix)
+
+hPutArray h arr sz   = hPutBufBA h arr sz
+hGetArray h sz       = hGetBufBA h sz
+
+mkIOError :: IOErrorType -> String -> Maybe Handle -> Maybe FilePath -> Exception
+mkIOError t location maybe_hdl maybe_filename
+  = IOException (IOError maybe_hdl t location ""
+                 maybe_filename
+        )
+
+eofErrorType = EOF
+
+#ifndef SIZEOF_HSINT
+#define SIZEOF_HSINT  INT_SIZE_IN_BYTES
+#endif
+
+#ifndef SIZEOF_HSWORD
+#define SIZEOF_HSWORD WORD_SIZE_IN_BYTES
+#endif
+
+#else
+type BinArray = IOUArray Int Word8
+#endif
+
+data BinHandle
+  = BinMem {        -- binary data stored in an unboxed array
+     off_r :: !FastMutInt,      -- the current offset
+     sz_r  :: !FastMutInt,      -- size of the array (cached)
+     arr_r :: !(IORef BinArray),    -- the array (bounds: (0,size-1))
+     bit_off_r :: !FastMutInt,          -- the bit offset (see end of file)
+     bit_cache_r :: !FastMutInt           -- the bit cache  (see end of file)
+    }
+    -- XXX: should really store a "high water mark" for dumping out
+    -- the binary data to a file.
+
+  | BinIO {     -- binary data stored in a file
+     off_r :: !FastMutInt,      -- the current offset (cached)
+     hdl   :: !IO.Handle,               -- the file handle (must be seekable)
+     bit_off_r :: !FastMutInt,          -- the bit offset (see end of file)
+     bit_cache_r :: !FastMutInt           -- the bit cache  (see end of file)
+   }
+    -- cache the file ptr in BinIO; using hTell is too expensive
+    -- to call repeatedly.  If anyone else is modifying this Handle
+    -- at the same time, we'll be screwed.
+
+data Bin a = BinPtr !Int !Int -- byte/bit
+  deriving (Eq, Ord, Show, Bounded)
+
+castBin :: Bin a -> Bin b
+castBin (BinPtr i j) = BinPtr i j
+
+class Binary a where
+    put_   :: BinHandle -> a -> IO ()
+    put    :: BinHandle -> a -> IO (Bin a)
+    get    :: BinHandle -> IO a
+
+    -- define one of put_, put.  Use of put_ is recommended because it
+    -- is more likely that tail-calls can kick in, and we rarely need the
+    -- position return value.
+    put_ bh a = do put bh a; return ()
+    put bh a  = do p <- tellBin bh; put_ bh a; return p
+
+putAt  :: Binary a => BinHandle -> Bin a -> a -> IO ()
+putAt bh p x = do seekBin bh p; put bh x; return ()
+
+getAt  :: Binary a => BinHandle -> Bin a -> IO a
+getAt bh p = do seekBin bh p; get bh
+
+openBinIO_ :: IO.Handle -> IO BinHandle
+openBinIO_ h = openBinIO h noBinHandleUserData
+
+newZeroInt = do r <- newFastMutInt; writeFastMutInt r 0; return r
+
+-- openBinIO :: IO.Handle -> Module -> IO BinHandle
+openBinIO :: forall t. Handle -> t -> IO BinHandle
+openBinIO h mod = do
+  r <- newZeroInt
+  o <- newZeroInt
+  c <- newZeroInt
+--  state <- newWriteState mod
+  return (BinIO r h o c)
+
+--openBinMem :: Int -> Module -> IO BinHandle
+openBinMem :: forall t. Int -> t -> IO BinHandle
+openBinMem size mod
+ | size <= 0 = error "Data.Binary.openBinMem: size must be > 0"   -- fix, was ">= 0"
+ | otherwise = do
+   arr <- newArray_ (0,size-1)
+   arr_r <- newIORef arr
+   ix_r <- newFastMutInt
+   writeFastMutInt ix_r 0
+   sz_r <- newFastMutInt
+   writeFastMutInt sz_r size
+   o <- newZeroInt
+   c <- newZeroInt
+--   state <- newWriteState mod
+   return (BinMem ix_r sz_r arr_r o c)
+
+noBinHandleUserData = error "Binary.BinHandle: no user data"
+
+--getUserData :: BinHandle -> BinHandleState
+--getUserData bh = state bh
+
+tellBin :: BinHandle -> IO (Bin a)
+tellBin (BinIO r _ o _)   =  do ix <- readFastMutInt r; bix <- readFastMutInt o; return (BinPtr ix bix)
+tellBin (BinMem r _ _ o _) = do ix <- readFastMutInt r; bix <- readFastMutInt o; return (BinPtr ix bix)
+
+tellBinByte (BinIO r _ _ _)    = do ix <- readFastMutInt r; return ix
+tellBinByte (BinMem r _ _ _ _) = do ix <- readFastMutInt r; return ix
+
+seekBin :: BinHandle -> Bin a -> IO ()
+seekBin bh@(BinIO ix_r h o c) (BinPtr p bit) = do 
+  writeFastMutInt ix_r p
+  writeFastMutInt o 0
+  writeFastMutInt c 0
+  hSeek h AbsoluteSeek (fromIntegral p)
+  when (bit /= 0) $ getBits bh bit >> return ()
+  return ()
+seekBin h@(BinMem ix_r sz_r a o c) (BinPtr p bit) = do
+  sz <- readFastMutInt sz_r
+  if (p >= sz)
+    then do expandBin h p
+            writeFastMutInt ix_r p
+            writeFastMutInt o 0
+            writeFastMutInt c 0
+            when (bit /= 0) $ getBits h bit >> return ()
+            return ()
+
+    else do writeFastMutInt ix_r p
+            writeFastMutInt o 0
+            writeFastMutInt c 0
+            when (bit /= 0) $ getBits h bit >> return ()
+            return ()
+
+isEOFBin :: BinHandle -> IO Bool
+isEOFBin (BinMem ix_r sz_r a _ _) = do
+  ix <- readFastMutInt ix_r
+  sz <- readFastMutInt sz_r
+  return (ix >= sz)
+isEOFBin (BinIO ix_r h _ _) = hIsEOF h
+
+writeBinMem :: BinHandle -> FilePath -> IO ()
+writeBinMem (BinIO _ _ _ _) _ = error "Data.Binary.writeBinMem: not a memory handle"
+writeBinMem bh@(BinMem ix_r sz_r arr_r bit_off_r bit_cache_r) fn = do
+  flushByte bh
+  h <- openBinaryFile fn WriteMode
+  arr <- readIORef arr_r
+  ix  <- readFastMutInt ix_r
+  hPutArray h arr ix
+  hClose h
+
+flushByte :: BinHandle -> IO ()
+flushByte bh = do
+  bit_off <- readFastMutInt (bit_off_r bh)
+  if bit_off == 0
+    then return ()
+    else putBits bh (8 - bit_off) 0
+
+finishByte :: BinHandle -> IO ()
+finishByte bh = do
+  bit_off <- readFastMutInt (bit_off_r bh)
+  if bit_off == 0
+    then return ()
+    else getBits bh (8 - bit_off) >> return ()
+
+readBinMem :: FilePath -> IO BinHandle
+readBinMem filename = do
+  h <- openBinaryFile filename ReadMode
+  filesize' <- hFileSize h
+  let filesize = fromIntegral filesize'
+  arr <- newArray_ (0,filesize-1)
+  count <- hGetArray h arr filesize
+  when (count /= filesize)
+    (error ("Binary.readBinMem: only read " ++ show count ++ " bytes"))
+  hClose h
+  arr_r <- newIORef arr
+  ix_r <- newFastMutInt
+  writeFastMutInt ix_r 0
+  sz_r <- newFastMutInt
+  writeFastMutInt sz_r filesize
+  bit_off_r <- newZeroInt
+  bit_cache_r <- newZeroInt
+  return (BinMem {-initReadState-} ix_r sz_r arr_r bit_off_r bit_cache_r)
+
+-- expand the size of the array to include a specified offset
+expandBin :: BinHandle -> Int -> IO ()
+expandBin (BinMem ix_r sz_r arr_r _ _) off = do
+   sz <- readFastMutInt sz_r
+   let sz' = head (dropWhile (<= off) (iterate (* 2) sz))
+   arr <- readIORef arr_r
+   arr' <- newArray_ (0,sz'-1)
+   sequence_ [ unsafeRead arr i >>= unsafeWrite arr' i
+         | i <- [ 0 .. sz-1 ] ]
+   writeFastMutInt sz_r sz'
+   writeIORef arr_r arr'
+--   hPutStrLn stderr ("expanding to size: " ++ show sz')
+   return ()
+expandBin (BinIO _ _ _ _) _ = return ()
+    -- no need to expand a file, we'll assume they expand by themselves.
+
+-- -----------------------------------------------------------------------------
+-- Low-level reading/writing of bytes
+
+putWord8 :: BinHandle -> Word8 -> IO ()
+putWord8 h@(BinMem ix_r sz_r arr_r bit_off_r bit_cache_r) w = do
+    bit_off <- readFastMutInt bit_off_r
+    if bit_off /= 0 then putBits h 8 w else do   -- only do standard putWord8 if bit_off == 0
+    ix <- readFastMutInt ix_r
+    sz <- readFastMutInt sz_r
+    -- double the size of the array if it overflows
+    if (ix >= sz) 
+        then do expandBin h ix
+                putWord8 h w
+        else do arr <- readIORef arr_r
+                unsafeWrite arr ix w
+                writeFastMutInt ix_r (ix+1)
+                return ()
+
+putWord8 bh@(BinIO ix_r h bit_off_r bit_cache_r) w = do
+    bit_off <- readFastMutInt bit_off_r
+    if bit_off /= 0 then putBits bh 8 w else do
+        ix <- readFastMutInt ix_r
+        hPutChar h (chr (fromIntegral w))   -- XXX not really correct
+        writeFastMutInt ix_r (ix+1)
+        return ()
+
+putByteNoBits :: BinHandle -> Word8 -> IO ()
+putByteNoBits h@(BinMem ix_r sz_r arr_r _ _) w = do
+    ix <- readFastMutInt ix_r
+    sz <- readFastMutInt sz_r
+    -- double the size of the array if it overflows
+    if (ix >= sz) 
+        then do expandBin h ix
+                putByteNoBits h w
+        else do arr <- readIORef arr_r
+                unsafeWrite arr ix w
+                writeFastMutInt ix_r (ix+1)
+                return ()
+
+putByteNoBits bh@(BinIO ix_r h _ _) w = do
+    hPutChar h (chr (fromIntegral w))   -- XXX not really correct
+    incFastMutInt ix_r
+    return ()
+
+getByteNoBits :: BinHandle -> IO Word8
+getByteNoBits h@(BinMem ix_r sz_r arr_r _ _) = do
+    ix <- readFastMutInt ix_r
+    sz <- readFastMutInt sz_r
+    when (ix >= sz)  $
+        throw (IOException $ mkIOError eofErrorType "Data.Binary.getWord8" Nothing Nothing)
+    arr <- readIORef arr_r
+    w <- unsafeRead arr ix
+    writeFastMutInt ix_r (ix+1)
+    return w
+
+getByteNoBits bh@(BinIO ix_r h _ _) = do
+    c <- hGetChar h
+    incFastMutInt ix_r
+    return $! (fromIntegral (ord c))    -- XXX not really correct
+
+getWord8 :: BinHandle -> IO Word8
+getWord8 h@(BinMem ix_r sz_r arr_r bit_off_r _) = do
+    bit_off <- readFastMutInt bit_off_r
+    if bit_off /= 0 then getBits h 8 else do
+    ix <- readFastMutInt ix_r
+    sz <- readFastMutInt sz_r
+    when (ix >= sz)  $
+        throw (IOException $ mkIOError eofErrorType "Data.Binary.getWord8" Nothing Nothing)
+    arr <- readIORef arr_r
+    w <- unsafeRead arr ix
+    writeFastMutInt ix_r (ix+1)
+    return w
+getWord8 bh@(BinIO ix_r h bit_off_r _) = do
+    bit_off <- readFastMutInt bit_off_r
+    if bit_off /= 0 then getBits bh 8 else do
+    ix <- readFastMutInt ix_r
+    c <- hGetChar h
+    writeFastMutInt ix_r (ix+1)
+    return $! (fromIntegral (ord c))    -- XXX not really correct
+
+putByte :: BinHandle -> Word8 -> IO ()
+putByte bh w = put_ bh w
+
+getByte :: BinHandle -> IO Word8
+getByte = getWord8
+
+-- -----------------------------------------------------------------------------
+-- Bit functions
+
+putBits :: BinHandle -> Int -> Word8 -> IO ()
+putBits bh num_bits bits {- | num_bits == 0 = return ()
+                         | num_bits <  0 = error "putBits cannot write negative numbers of bits"
+                         | num_bits >  8 = error "putBits cannot write more than 8 bits at a time"
+                         | otherwise    -} = do
+  bit_off <- readFastMutInt (bit_off_r bh)
+  if num_bits + bit_off < 8
+    then do incFastMutIntBy (bit_off_r bh) num_bits
+            orFastMutInt (bit_cache_r bh) (bits `shiftL` bit_off)
+    else if num_bits + bit_off == 8
+           then do writeFastMutInt (bit_off_r bh) 0
+                   bit_cache <- {-# SCC "bc1" #-} readFastMutInt (bit_cache_r bh) >>= return . fromIntegral
+                   writeFastMutInt (bit_cache_r bh) 0
+                   --putByte bh (bit_cache .|. (bits `shiftL` bit_off))    -- won't call putBits because bit_off_r == 0
+                   putByteNoBits bh (bit_cache .|. (bits `shiftL` bit_off))
+
+           else do let leftover_bits = 8 - bit_off                       -- we are going over a byte boundary
+                   bit_cache <- {-# SCC "bc2" #-} readFastMutInt (bit_cache_r bh) >>= \x -> return ({-# SCC "fi" #-} fromIntegral x)
+                   writeFastMutInt (bit_off_r bh) 0
+                   writeFastMutInt (bit_cache_r bh) 0
+                   {- putByte bh (bit_cache .|. (bits `shiftL` bit_off))  -}  -- won't call putBits
+                   putByteNoBits bh (bit_cache .|. (bits `shiftL` bit_off))
+                   putBits bh (num_bits - leftover_bits) (bits `shiftR` leftover_bits)
+
+getBits :: BinHandle -> Int -> IO Word8
+getBits bh num_bits {- | num_bits == 0 = return 0
+                    | num_bits <  0 = error "getBits cannot read negative numbers of bits"
+                    | num_bits >  8 = error "getBits cannot read more than 8 bits at a time"
+                    | otherwise     -} = do
+  bit_off <- readFastMutInt (bit_off_r bh)
+  if bit_off == 0
+    then do bit_cache <- getByte bh
+            if num_bits == 8
+              then do writeFastMutInt (bit_off_r   bh) 0
+                      writeFastMutInt (bit_cache_r bh) 0
+                      return bit_cache
+              else do writeFastMutInt (bit_off_r   bh) (fromIntegral num_bits)
+                      writeFastMutInt (bit_cache_r bh) (fromIntegral bit_cache)
+                      return (bit_cache .&. bit_mask num_bits)
+    else if bit_off + num_bits < 8
+    then do incFastMutIntBy (bit_off_r bh) num_bits
+            bit_cache <- readFastMutInt (bit_cache_r bh) >>= return . fromIntegral
+            return ((bit_cache `shiftR` bit_off) .&. bit_mask num_bits)
+    else if bit_off + num_bits == 8
+    then do writeFastMutInt (bit_off_r bh) 0
+            bit_cache <- readFastMutInt (bit_cache_r bh) >>= return . fromIntegral
+            writeFastMutInt (bit_cache_r bh) 0
+            return ((bit_cache `shiftR` bit_off) .&. bit_mask num_bits)
+    else do let leftover_bits = 8 - bit_off
+            bit_cache <- readFastMutInt (bit_cache_r bh) >>= return . fromIntegral
+            let bits = (bit_cache `shiftR` bit_off) .&. bit_mask leftover_bits
+            writeFastMutInt (bit_cache_r bh) 0
+            writeFastMutInt (bit_off_r   bh) 0
+            {- bit_cache <- getByte bh -}
+            -- use a version that doesn't care about bits
+            bit_cache <- getByteNoBits bh
+            writeFastMutInt (bit_off_r   bh) (num_bits - leftover_bits)
+            writeFastMutInt (bit_cache_r bh) (fromIntegral bit_cache)
+            return (bits .|. ((bit_cache .&. bit_mask (num_bits - leftover_bits)) `shiftL` leftover_bits))
+
+            
+bit_mask n = (complement 0) `shiftR` (8 - n)
+
+-- -----------------------------------------------------------------------------
+-- Primitve Word writes
+
+instance Binary Word8 where
+  put_ = putWord8
+  get  = getWord8
+
+instance Binary Word16 where
+  put_ h w = do -- XXX too slow.. inline putWord8?
+    putByte h (fromIntegral (w `shiftR` 8))
+    putByte h (fromIntegral (w .&. 0xff))
+  get h = do
+    w1 <- getWord8 h
+    w2 <- getWord8 h
+    return $! ((fromIntegral w1 `shiftL` 8) .|. fromIntegral w2)
+
+
+instance Binary Word32 where
+  put_ h w = do
+    putByte h (fromIntegral (w `shiftR` 24))
+    putByte h (fromIntegral ((w `shiftR` 16) .&. 0xff))
+    putByte h (fromIntegral ((w `shiftR` 8)  .&. 0xff))
+    putByte h (fromIntegral (w .&. 0xff))
+  get h = do
+    w1 <- getWord8 h
+    w2 <- getWord8 h
+    w3 <- getWord8 h
+    w4 <- getWord8 h
+    return $! ((fromIntegral w1 `shiftL` 24) .|. 
+           (fromIntegral w2 `shiftL` 16) .|. 
+           (fromIntegral w3 `shiftL`  8) .|. 
+           (fromIntegral w4))
+
+
+instance Binary Word64 where
+  put_ h w = do
+    putByte h (fromIntegral (w `shiftR` 56))
+    putByte h (fromIntegral ((w `shiftR` 48) .&. 0xff))
+    putByte h (fromIntegral ((w `shiftR` 40) .&. 0xff))
+    putByte h (fromIntegral ((w `shiftR` 32) .&. 0xff))
+    putByte h (fromIntegral ((w `shiftR` 24) .&. 0xff))
+    putByte h (fromIntegral ((w `shiftR` 16) .&. 0xff))
+    putByte h (fromIntegral ((w `shiftR`  8) .&. 0xff))
+    putByte h (fromIntegral (w .&. 0xff))
+  get h = do
+    w1 <- getWord8 h
+    w2 <- getWord8 h
+    w3 <- getWord8 h
+    w4 <- getWord8 h
+    w5 <- getWord8 h
+    w6 <- getWord8 h
+    w7 <- getWord8 h
+    w8 <- getWord8 h
+    return $! ((fromIntegral w1 `shiftL` 56) .|. 
+           (fromIntegral w2 `shiftL` 48) .|. 
+           (fromIntegral w3 `shiftL` 40) .|. 
+           (fromIntegral w4 `shiftL` 32) .|. 
+           (fromIntegral w5 `shiftL` 24) .|. 
+           (fromIntegral w6 `shiftL` 16) .|. 
+           (fromIntegral w7 `shiftL`  8) .|. 
+           (fromIntegral w8))
+
+-- -----------------------------------------------------------------------------
+-- Primitve Int writes
+
+instance Binary Int8 where
+  put_ h w = put_ h (fromIntegral w :: Word8)
+  get h    = do w <- get h; return $! (fromIntegral (w::Word8))
+
+instance Binary Int16 where
+  put_ h w = put_ h (fromIntegral w :: Word16)
+  get h    = do w <- get h; return $! (fromIntegral (w::Word16))
+
+instance Binary Int32 where
+  put_ h w = put_ h (fromIntegral w :: Word32)
+  get h    = do w <- get h; return $! (fromIntegral (w::Word32))
+
+put31ofInt32 :: BinHandle -> Int32 -> IO ()
+put31ofInt32 h i = do
+    putBits h 7 (fromIntegral (w `shiftR` 24))
+    putBits h 8 (fromIntegral ((w `shiftR` 16) .&. 0xff))
+    putBits h 8 (fromIntegral ((w `shiftR` 8)  .&. 0xff))
+    putBits h 8 (fromIntegral (w .&. 0xff))
+    where w = fromIntegral i :: Word32
+
+get31ofInt32 :: BinHandle -> IO Int32
+get31ofInt32 h = do
+    w1 <- getBits  h 7
+    w2 <- getWord8 h
+    w3 <- getWord8 h
+    w4 <- getWord8 h
+    return $! ((fromIntegral w1 `shiftL` 24) .|. 
+           (fromIntegral w2 `shiftL` 16) .|. 
+           (fromIntegral w3 `shiftL`  8) .|. 
+           (fromIntegral w4))
+
+instance Binary Int64 where
+  put_ h w = put_ h (fromIntegral w :: Word64)
+  get h    = do w <- get h; return $! (fromIntegral (w::Word64))
+
+-- -----------------------------------------------------------------------------
+-- Instances for standard types
+
+instance Binary () where
+    put_ bh () = return ()
+    get  _     = return ()
+--    getF bh p  = case getBitsF bh 0 p of (_,b) -> ((),b)
+
+{- updated for bits
+instance Binary Bool where
+    put_ bh b = putByte bh (fromIntegral (fromEnum b))
+    get  bh   = do x <- getWord8 bh; return $! (toEnum (fromIntegral x))
+--    getF bh p = case getBitsF bh 1 p of (x,b) -> (toEnum x,b)
+-}
+
+instance Binary Bool where
+    put_ bh True  = putBits bh 1 1
+    put_ bh False = putBits bh 1 0
+    get  bh = do b <- getBits bh 1; return (b == 1)
+
+instance Binary Char where
+    put_  bh c = put_ bh (fromIntegral (ord c) :: Word32)
+    get  bh   = do x <- get bh; return $! (chr (fromIntegral (x :: Word32)))
+--    getF bh p = case getBitsF bh 8 p of (x,b) -> (toEnum x,b)
+
+instance Binary Int where
+#if SIZEOF_HSINT == 4
+    put_ bh i = put_ bh (fromIntegral i :: Int32)
+    get  bh = do
+    x <- get bh
+    return $! (fromIntegral (x :: Int32))
+#elif SIZEOF_HSINT == 8
+    put_ bh i = put_ bh (fromIntegral i :: Int64)
+    get  bh = do
+    x <- get bh
+    return $! (fromIntegral (x :: Int64))
+#else
+#error "unsupported sizeof(HsInt)"
+#endif
+--    getF bh   = getBitsF bh 32
+
+{-
+instance Binary a => Binary [a] where
+    put_ bh []     = putByte bh 0
+    put_ bh (x:xs) = do putByte bh 1; put_ bh x; put_ bh xs
+    get bh         = do h <- getWord8 bh
+                        case h of
+                          0 -> return []
+                          _ -> do x  <- get bh
+                                  xs <- get bh
+                                  return (x:xs)
+-}
+
+instance Binary a => Binary [a] where
+    put_ bh l = do
+       put_ bh (length l)
+       mapM (put_ bh) l
+       return ()
+    get bh = do
+       len <- get bh
+       mapM (\_ -> get bh) [1..(len::Int)]
+
+instance (Binary a, Binary b) => Binary (a,b) where
+    put_ bh (a,b) = do put_ bh a; put_ bh b
+    get bh        = do a <- get bh
+                       b <- get bh
+                       return (a,b)
+
+instance (Binary a, Binary b, Binary c) => Binary (a,b,c) where
+    put_ bh (a,b,c) = do put_ bh a; put_ bh b; put_ bh c
+    get bh          = do a <- get bh
+                         b <- get bh
+                         c <- get bh
+                         return (a,b,c)
+
+instance (Binary a, Binary b, Binary c, Binary d) => Binary (a,b,c,d) where
+    put_ bh (a,b,c,d) = do put_ bh a; put_ bh b; put_ bh c; put_ bh d
+    get bh          = do a <- get bh
+                         b <- get bh
+                         c <- get bh
+                         d <- get bh
+                         return (a,b,c,d)
+
+instance (Binary a, Binary b, Binary c, Binary d, Binary e) => Binary (a,b,c,d,e) where
+    put_ bh (a,b,c,d,e) = do put_ bh a; put_ bh b; put_ bh c; put_ bh d; put_ bh e
+    get bh          = do a <- get bh
+                         b <- get bh
+                         c <- get bh
+                         d <- get bh
+                         e <- get bh
+                         return (a,b,c,d,e)
+
+instance (Binary a, Binary b, Binary c, Binary d, Binary e, Binary f) => Binary (a,b,c,d,e,f) where
+    put_ bh (a,b,c,d,e,f) = do put_ bh a; put_ bh b; put_ bh c; put_ bh d; put_ bh e; put_ bh f
+    get bh          = do a <- get bh
+                         b <- get bh
+                         c <- get bh
+                         d <- get bh
+                         e <- get bh
+                         f <- get bh
+                         return (a,b,c,d,e,f)
+
+instance Binary a => Binary (Maybe a) where
+    put_ bh Nothing  = putByte bh 0
+    put_ bh (Just a) = do putByte bh 1; put_ bh a
+    get bh           = do h <- getWord8 bh
+                          case h of
+                            0 -> return Nothing
+                            _ -> do x <- get bh; return (Just x)
+
+putMaybeInt :: BinHandle -> Maybe Int -> IO ()
+getMaybeInt :: BinHandle -> IO (Maybe Int)
+putMaybeInt bh Nothing = putBits bh 1 0
+putMaybeInt bh (Just i) = do putBits bh 1 1; put31ofInt32 bh (fromIntegral i)
+
+getMaybeInt bh = do 
+  b <- getBits bh 1
+  case b of
+    0 -> return Nothing
+    _ -> do i <- get31ofInt32 bh
+            return (Just (fromIntegral i))
+
+{- RULES get = getMaybeInt -}
+
+{- SPECIALIZE put_ :: BinHandle -> Maybe Int -> IO () = putMaybeInt -}
+{- SPECIALIZE get  :: BinHandle -> IO (Maybe Int)     = getMaybeInt -}
+
+
+instance (Binary a, Binary b) => Binary (Either a b) where
+    put_ bh (Left  a) = do putByte bh 0; put_ bh a
+    put_ bh (Right b) = do putByte bh 1; put_ bh b
+    get bh            = do h <- getWord8 bh
+                           case h of
+                             0 -> do a <- get bh ; return (Left a)
+                             _ -> do b <- get bh ; return (Right b)
+
+instance Binary Integer where
+    put_ bh (S# i#) = do putByte bh 0; put_ bh (I# i#)
+    put_ bh (J# s# a#) = do
+        p <- putByte bh 1;
+        put_ bh (I# s#)
+        let sz# = sizeofByteArray# a#  -- in *bytes*
+        put_ bh (I# sz#)  -- in *bytes*
+        putByteArray bh a# sz#
+
+    get bh = do
+        b <- getByte bh
+        case b of
+          0 -> do (I# i#) <- get bh
+                  return (S# i#)
+          _ -> do (I# s#) <- get bh
+                  sz <- get bh
+                  (BA a#) <- getByteArray bh sz
+                  return (J# s# a#)
+
+putByteArray :: BinHandle -> ByteArray# -> Int# -> IO ()
+putByteArray bh a s# = loop 0#
+  where loop n# 
+           | n# ==# s# = return ()
+           | otherwise = do
+                putByte bh (indexByteArray a n#)
+                loop (n# +# 1#)
+
+getByteArray :: BinHandle -> Int -> IO ByteArray
+getByteArray bh (I# sz) = do
+  (MBA arr) <- newByteArray sz 
+  let loop n
+       | n ==# sz = return ()
+       | otherwise = do
+        w <- getByte bh 
+        writeByteArray arr n w
+        loop (n +# 1#)
+  loop 0#
+  freezeByteArray arr
+
+
+data ByteArray = BA ByteArray#
+data MBA = MBA (MutableByteArray# RealWorld)
+
+newByteArray :: Int# -> IO MBA
+newByteArray sz = IO $ \s ->
+  case newByteArray# sz s of { (# s, arr #) ->
+  (# s, MBA arr #) }
+
+freezeByteArray :: MutableByteArray# RealWorld -> IO ByteArray
+freezeByteArray arr = IO $ \s ->
+  case unsafeFreezeByteArray# arr s of { (# s, arr #) ->
+  (# s, BA arr #) }
+
+writeByteArray :: MutableByteArray# RealWorld -> Int# -> Word8 -> IO ()
+
+writeByteArray arr i w8 = IO $ \s ->
+  case fromIntegral w8 of { W# w# -> 
+  case writeCharArray# arr i (chr# (word2Int# w#)) s  of { s ->
+  (# s , () #) }}
+
+indexByteArray a# n# = fromIntegral (I# (ord# (indexCharArray# a# n#)))
+
+instance (Integral a, Binary a) => Binary (Ratio a) where
+    put_ bh (a :% b) = do put_ bh a; put_ bh b
+    get bh = do a <- get bh; b <- get bh; return (a :% b)
+
+instance Binary (Bin a) where
+  put_ bh (BinPtr i j) = put_ bh (i,j)
+  get bh = do (i,j) <- get bh; return (BinPtr i j)
+
+-- -----------------------------------------------------------------------------
+-- Lazy reading/writing
+
+lazyPut :: Binary a => BinHandle -> a -> IO ()
+lazyPut bh a = do
+    -- output the obj with a ptr to skip over it:
+    pre_a <- tellBin bh
+    put_ bh pre_a   -- save a slot for the ptr
+    put_ bh a       -- dump the object
+    q <- tellBin bh     -- q = ptr to after object
+    putAt bh pre_a q    -- fill in slot before a with ptr to q
+    seekBin bh q    -- finally carry on writing at q
+
+lazyGet :: Binary a => BinHandle -> IO a
+lazyGet bh = do
+    p <- get bh     -- a BinPtr
+    p_a <- tellBin bh
+    a <- unsafeInterleaveIO (getAt bh p_a)
+    seekBin bh p -- skip over the object for now
+    return a
+
+-- -----------------------------------------------------------------------------
+-- BinHandleState
+{-
+type BinHandleState = 
+    (Module, 
+     IORef Int,
+     IORef (UniqFM (Int,FastString)),
+     Array Int FastString)
+
+initReadState :: BinHandleState
+initReadState = (undef, undef, undef, undef)
+
+newWriteState :: Module -> IO BinHandleState
+newWriteState m = do
+  j_r <- newIORef 0
+  out_r <- newIORef emptyUFM
+  return (m,j_r,out_r,undef)
+
+undef = error "Binary.BinHandleState"
+
+-- -----------------------------------------------------------------------------
+-- FastString binary interface
+
+getBinFileWithDict :: Binary a => FilePath -> IO a
+getBinFileWithDict file_path = do
+  bh <- Binary.readBinMem file_path
+  magic <- get bh
+  when (magic /= binaryInterfaceMagic) $
+    throwDyn (ProgramError (
+       "magic number mismatch: old/corrupt interface file?"))
+  dict_p <- Binary.get bh       -- get the dictionary ptr
+  data_p <- tellBin bh
+  seekBin bh dict_p
+  dict <- getDictionary bh
+  seekBin bh data_p
+  let (mod, j_r, out_r, _) = state bh
+  get bh{ state = (mod,j_r,out_r,dict) }
+
+initBinMemSize = (1024*1024) :: Int
+
+binaryInterfaceMagic = 0x1face :: Word32
+
+putBinFileWithDict :: Binary a => FilePath -> Module -> a -> IO ()
+putBinFileWithDict file_path mod a = do
+  bh <- openBinMem initBinMemSize mod
+  put_ bh binaryInterfaceMagic
+  p <- tellBin bh
+  put_ bh p     -- placeholder for ptr to dictionary
+  put_ bh a
+  let (_, j_r, fm_r, _) = state bh
+  j <- readIORef j_r
+  fm <- readIORef fm_r
+  dict_p <- tellBin bh
+  putAt bh p dict_p -- fill in the placeholder
+  seekBin bh dict_p -- seek back to the end of the file
+  putDictionary bh j (constructDictionary j fm)
+  writeBinMem bh file_path
+  
+type Dictionary = Array Int FastString
+    -- should be 0-indexed
+
+putDictionary :: BinHandle -> Int -> Dictionary -> IO ()
+putDictionary bh sz dict = do
+  put_ bh sz
+  mapM_ (putFS bh) (elems dict)
+
+getDictionary :: BinHandle -> IO Dictionary
+getDictionary bh = do 
+  sz <- get bh
+  elems <- sequence (take sz (repeat (getFS bh)))
+  return (listArray (0,sz-1) elems)
+
+constructDictionary :: Int -> UniqFM (Int,FastString) -> Dictionary
+constructDictionary j fm = array (0,j-1) (eltsUFM fm)
+
+putFS bh (FastString id l ba) = do
+  put_ bh (I# l)
+  putByteArray bh ba l
+putFS bh s = error ("Binary.put_(FastString): " ++ unpackFS s)
+    -- Note: the length of the FastString is *not* the same as
+    -- the size of the ByteArray: the latter is rounded up to a
+    -- multiple of the word size.
+  
+{- -- possible faster version, not quite there yet:
+getFS bh@BinMem{} = do
+  (I# l) <- get bh
+  arr <- readIORef (arr_r bh)
+  off <- readFastMutInt (off_r bh)
+  return $! (mkFastSubStringBA# arr off l)
+-}
+getFS bh = do
+  (I# l) <- get bh
+  (BA ba) <- getByteArray bh (I# l)
+  return $! (mkFastSubStringBA# ba 0# l)
+
+instance Binary FastString where
+  put_ bh f@(FastString id l ba) =
+    case getUserData bh of { (_, j_r, out_r, dict) -> do
+    out <- readIORef out_r
+    let uniq = getUnique f
+    case lookupUFM out uniq of
+    Just (j,f)  -> put_ bh j
+    Nothing -> do
+       j <- readIORef j_r
+       put_ bh j
+       writeIORef j_r (j+1)
+       writeIORef out_r (addToUFM out uniq (j,f))
+    }
+  put_ bh s = error ("Binary.put_(FastString): " ++ show (unpackFS s))
+
+  get bh = do 
+    j <- get bh
+    case getUserData bh of (_, _, _, arr) -> return (arr ! j)
+-}
+
+
+
+{----------------------------------------------------------------------
+ ---------- Hal's Notes -----------------------------------------------
+ ----------------------------------------------------------------------
+
+We are adding support for 
+
+  putBits   :: BinHandle -> Int -> Word8 -> IO ()
+  getBits   :: BinHandle -> Int -> IO Word8
+  flushBits :: BinHandle -> Int -> IO ()
+  closeHandle :: BinHandle -> IO ()
+
+where
+
+  `putBits bh num_bits bits' writes the right-most num_bits of bits to
+  bh.  `getBits bh num_bits` reads num_bits from bh and stores them in
+  the right-most positions of the result.  flushBits bh n alignes the
+  stream to the next 2^n bit boundary.  closeHandle flushes all
+  remaining bits and closes the handle.
+
+In order to implement this, we need to extend the BinHandles with two
+fields: bit_off_r :: Int and bit_cache :: Word8.  Based on this, the
+basic implementations look something like this:
+
+putBits bh num_bits bits =
+  if num_bits + bit_off_r <= 8
+    then bit_off_r += num_bits
+         add num_bits of bits to the tail of bit_cache
+         if bit_off_r == 8
+           then write bit_cache and set bit_cache = 0, bit_off_r = 0
+    else let leftover_bits = 8 - bit_off_r
+         add leftover_bits of bits to tail of bit_cache
+         write bit_cache and set bit_cache = 0, bit_off_r = 0
+         putBits bh (num_bits - leftover_bits) (bits >> leftover_bits)
+
+(note that this will recurse at most once)
+
+getBits bh num_bits =
+  if bit_off_r == 0
+    then bit_cache <- read a byte
+         bit_off_r = num_bits
+         if bit_off_r == 8, set bit_off_r = 0, bit_cache = 0
+    else if bit_off_r + num_bits <= 8
+           then bit_off_r += num_bits
+                bits = bits from bit_off_r -> bit_off_r+num_bits of bit_cache
+                if bit_off_r == 8, set bit_off_r = 0, bit_cache = 0
+                return bits
+           else let leftover_bits = 8 - bit_off_r
+                bits = (last leftover_bits from bit_cache) << (num_bits - leftover_bits)
+                bit_cache <- read a byte
+                bit_off_r = num_bits - leftover_bits
+                return (bits || first (num_bits - leftover_bits) of bit_cache)
+
+Now, we must also modify putByte/getByte.  In these, we do a quick
+check to see if bit_off_r == 0; if it does, then we just execute
+normally.  Otherwise, we just call putBits/getBits with num_bits=8.
+
+closeHandle bh =
+  if bit_off_r == 0
+    then close the handle
+    else write bit_cache and set bit_cache = 0, bit_off_r =0
+         close the handle
+
+-}
+
+------------------------------------------------------------------------
+
+#if __GLASGOW_HASKELL__ < 411
+newByteArray# = newCharArray#
+#endif
+
+#ifdef __GLASGOW_HASKELL__
+
+data FastMutInt = FastMutInt (MutableByteArray# RealWorld)
+
+newFastMutInt :: IO FastMutInt
+newFastMutInt = IO $ \s ->
+  case newByteArray# size s of { (# s, arr #) ->
+  (# s, FastMutInt arr #) }
+  where I# size = SIZEOF_HSINT
+
+readFastMutInt :: FastMutInt -> IO Int
+readFastMutInt (FastMutInt arr) = IO $ \s ->
+  case readIntArray# arr 0# s of { (# s, i #) ->
+  (# s, I# i #) }
+
+writeFastMutInt :: FastMutInt -> Int -> IO ()
+writeFastMutInt (FastMutInt arr) (I# i) = IO $ \s ->
+  case writeIntArray# arr 0# i s of { s ->
+  (# s, () #) }
+
+incFastMutInt :: FastMutInt -> IO Int   -- Returns original value
+incFastMutInt (FastMutInt arr) = IO $ \s ->
+  case readIntArray# arr 0# s of { (# s, i #) ->
+  case writeIntArray# arr 0# (i +# 1#) s of { s ->
+  (# s, I# i #) } }
+
+incFastMutIntBy :: FastMutInt -> Int -> IO Int  -- Returns original value
+incFastMutIntBy (FastMutInt arr) (I# n) = IO $ \s ->
+  case readIntArray# arr 0# s of { (# s, i #) ->
+  case writeIntArray# arr 0# (i +# n) s of { s ->
+  (# s, I# i #) } }
+
+-- we should optimize this: ask SimonM :)
+orFastMutInt :: FastMutInt -> Word8 -> IO ()
+orFastMutInt fmi w = do
+  i <- readFastMutInt fmi
+  writeFastMutInt fmi (i .|. (fromIntegral w))
+
+#endif
+
diff --git a/tests/Parallel.hs b/tests/Parallel.hs
new file mode 100644
--- /dev/null
+++ b/tests/Parallel.hs
@@ -0,0 +1,147 @@
+-----------------------------------------------------------------------------
+-- |
+-- Module      :  Test.QuickCheck.Parallel
+-- Copyright   :  (c) Don Stewart 2006
+-- License     :  BSD-style (see the file LICENSE)
+-- 
+-- Maintainer  :  dons@cse.unsw.edu.au
+-- Stability   :  experimental
+-- Portability :  non-portable (uses Control.Exception, Control.Concurrent)
+--
+-- A parallel batch driver for running QuickCheck on threaded or SMP systems.
+-- See the /Example.hs/ file for a complete overview.
+--
+
+module Parallel (
+    pRun,
+    pDet,
+    pNon
+  ) where
+
+import Test.QuickCheck
+import Data.List
+import Control.Concurrent
+import Control.Exception  hiding (evaluate)
+import System.Random
+import System.IO          (hFlush,stdout)
+import Text.Printf
+
+type Name   = String
+type Depth  = Int
+type Test   = (Name, Depth -> IO String)
+
+-- | Run a list of QuickCheck properties in parallel chunks, using
+-- 'n' Haskell threads (first argument), and test to a depth of 'd'
+-- (second argument). Compile your application with '-threaded' and run
+-- with the SMP runtime's '-N4' (or however many OS threads you want to
+-- donate), for best results.
+--
+-- > import Test.QuickCheck.Parallel
+-- >
+-- > do n <- getArgs >>= readIO . head
+-- >    pRun n 1000 [ ("sort1", pDet prop_sort1) ]
+--
+-- Will run 'n' threads over the property list, to depth 1000.
+--
+pRun :: Int -> Int -> [Test] -> IO ()
+pRun n depth tests = do
+    chan <- newChan
+    ps   <- getChanContents chan
+    work <- newMVar tests
+
+    forM_ [1..n] $ forkIO . thread work chan
+
+    let wait xs i
+            | i >= n     = return () -- done
+            | otherwise = case xs of
+                    Nothing : xs -> wait xs $! i+1
+                    Just s  : xs -> putStr s >> hFlush stdout >> wait xs i
+    wait ps 0
+
+  where
+    thread :: MVar [Test] -> Chan (Maybe String) -> Int -> IO ()
+    thread work chan me = loop
+      where
+        loop = do
+            job <- modifyMVar work $ \jobs -> return $ case jobs of
+                        []     -> ([], Nothing)
+                        (j:js) -> (js, Just j)
+            case job of
+                Nothing          -> writeChan chan Nothing -- done
+                Just (name,prop) -> do
+                    v <- prop depth
+                    writeChan chan . Just $ printf "%d: %-25s: %s" me name v
+                    loop
+
+
+-- | Wrap a property, and run it on a deterministic set of data
+pDet :: Testable a => a -> Int -> IO String
+pDet a n = mycheck Det defaultConfig
+    { configMaxTest = n
+    , configEvery   = \n args -> unlines args } a
+
+-- | Wrap a property, and run it on a non-deterministic set of data
+pNon :: Testable a => a -> Int -> IO String
+pNon a n = mycheck NonDet defaultConfig
+    { configMaxTest = n
+    , configEvery   = \n args -> unlines args } a
+
+data Mode = Det | NonDet
+
+------------------------------------------------------------------------
+
+mycheck :: Testable a => Mode -> Config -> a -> IO String
+mycheck Det config a = do
+     let rnd = mkStdGen 99  -- deterministic
+     mytests config (evaluate a) rnd 0 0 []
+
+mycheck NonDet config a = do
+    rnd <- newStdGen        -- different each run
+    mytests config (evaluate a) rnd 0 0 []
+
+mytests :: Config -> Gen Result -> StdGen -> Int -> Int -> [[String]] -> IO String
+mytests config gen rnd0 ntest nfail stamps
+  | ntest == configMaxTest config = do done "OK," ntest stamps
+  | nfail == configMaxFail config = do done "Arguments exhausted after" ntest stamps
+  | otherwise = do
+         case ok result of
+           Nothing    ->
+             mytests config gen rnd1 ntest (nfail+1) stamps
+           Just True  ->
+             mytests config gen rnd1 (ntest+1) nfail (stamp result:stamps)
+           Just False ->
+             return ( "Falsifiable after "
+                   ++ show ntest
+                   ++ " tests:\n"
+                   ++ unlines (arguments result)
+                    )
+     where
+      result      = generate (configSize config ntest) rnd2 gen
+      (rnd1,rnd2) = split rnd0
+
+done :: String -> Int -> [[String]] -> IO String
+done mesg ntest stamps =
+    return ( mesg ++ " " ++ show ntest ++ " tests" ++ table )
+  where
+    table = display
+        . map entry
+        . reverse
+        . sort
+        . map pairLength
+        . group
+        . sort
+        . filter (not . null)
+        $ stamps
+
+    display []  = ".\n"
+    display [x] = " (" ++ x ++ ").\n"
+    display xs  = ".\n" ++ unlines (map (++ ".") xs)
+
+    pairLength xss@(xs:_) = (length xss, xs)
+    entry (n, xs)         = percentage n ntest
+                          ++ " "
+                          ++ concat (intersperse ", " xs)
+
+    percentage n m        = show ((100 * n) `div` m) ++ "%"
+
+forM_ = flip mapM_
diff --git a/tests/QC.hs b/tests/QC.hs
new file mode 100644
--- /dev/null
+++ b/tests/QC.hs
@@ -0,0 +1,244 @@
+{-# OPTIONS_GHC -fglasgow-exts #-}
+module Main where
+
+import Data.Binary
+import Data.Binary.Put
+import Data.Binary.Get
+
+import Parallel
+
+import qualified Data.ByteString as B
+import qualified Data.ByteString.Internal as B
+import qualified Data.ByteString.Unsafe as B
+import qualified Data.ByteString.Lazy as L
+import qualified Data.ByteString.Lazy.Internal as L
+import qualified Data.Map as Map
+import qualified Data.Set as Set
+import qualified Data.IntMap as IntMap
+import qualified Data.IntSet as IntSet
+
+import Data.Array (Array)
+import Data.Array.IArray
+import Data.Array.Unboxed (UArray)
+
+import qualified Control.OldException as C (catch,evaluate)
+import Control.Monad
+import Foreign
+import System.Environment
+import System.IO
+import System.IO.Unsafe
+
+import Test.QuickCheck hiding (test)
+import QuickCheckUtils
+import Text.Printf
+
+-- import qualified Data.Sequence as Seq
+
+------------------------------------------------------------------------
+
+roundTrip :: (Eq a, Binary a) => a -> (L.ByteString -> L.ByteString) -> Bool
+roundTrip a f = a ==
+    {-# SCC "decode.refragment.encode" #-} decode (f (encode a))
+
+roundTripWith put get x =
+    forAll positiveList $ \xs ->
+    x == runGet get (refragment xs (runPut (put x)))
+
+-- make sure that a test fails
+errorish :: B a
+errorish a = unsafePerformIO $
+    C.catch (do C.evaluate a
+                return False)
+            (\_ -> return True)
+
+-- low level ones:
+
+prop_Word16be = roundTripWith putWord16be getWord16be
+prop_Word16le = roundTripWith putWord16le getWord16le
+prop_Word16host = roundTripWith putWord16host getWord16host
+
+prop_Word32be = roundTripWith putWord32be getWord32be
+prop_Word32le = roundTripWith putWord32le getWord32le
+prop_Word32host = roundTripWith putWord32host getWord32host
+
+prop_Word64be = roundTripWith putWord64be getWord64be
+prop_Word64le = roundTripWith putWord64le getWord64le
+prop_Word64host = roundTripWith putWord64host getWord64host
+
+prop_Wordhost = roundTripWith putWordhost getWordhost
+
+-- read too much:
+
+prop_bookworm x = errorish $ x == a && x /= b
+  where
+    (a,b) = decode (encode x)
+
+-- sanity:
+
+invariant_lbs :: L.ByteString -> Bool
+invariant_lbs (L.Empty)      = True
+invariant_lbs (L.Chunk x xs) = not (B.null x) && invariant_lbs xs
+
+prop_invariant :: (Binary a) => a -> Bool
+prop_invariant = invariant_lbs . encode
+
+-- be lazy!
+
+-- doesn't do fair testing of lazy put/get.
+-- tons of untested cases
+
+-- lazyTrip :: (Binary a, Eq a) => a -> Property
+-- lazyTrip a = forAll positiveList $ \xs ->
+--     a == (runGet lazyGet . refragment xs . runPut . lazyPut $ a)
+
+-- refragment a lazy bytestring's chunks
+refragment :: [Int] -> L.ByteString -> L.ByteString
+refragment [] lps = lps
+refragment (x:xs) lps =
+    let x' = fromIntegral . (+1) . abs $ x
+        rest = refragment xs (L.drop x' lps) in
+    L.append (L.fromChunks [B.concat . L.toChunks . L.take x' $ lps]) rest
+
+-- check identity of refragmentation
+prop_refragment lps xs = lps == refragment xs lps
+
+-- check that refragmention still hold invariant
+prop_refragment_inv lps xs = invariant_lbs $ refragment xs lps
+
+main :: IO ()
+main = do
+    hSetBuffering stdout NoBuffering
+    s <- getArgs
+    let x = if null s then 100 else read (head s)
+    pRun 2 x tests
+
+{-
+run :: [(String, Int -> IO ())] -> IO ()
+run tests = do
+    x <- getArgs
+    let n = if null x then 100 else read . head $ x
+    mapM_ (\(s,a) -> printf "%-50s" s >> a n) tests
+-}
+
+------------------------------------------------------------------------
+
+type T a = a -> Property
+type B a = a -> Bool
+
+p       :: Testable a => a -> Int -> IO String
+p       = pNon
+
+test    :: (Eq a, Binary a) => a -> Property
+test a  = forAll positiveList (roundTrip a . refragment)
+
+positiveList :: Gen [Int]
+positiveList = fmap (filter (/=0) . map abs) $ arbitrary
+
+-- tests :: [(String, Int -> IO String)]
+tests =
+-- utils
+        [ ("refragment id",        p prop_refragment     )
+        , ("refragment invariant", p prop_refragment_inv )
+
+-- boundaries
+        , ("read to much",  p (prop_bookworm :: B Word8     ))
+
+-- Primitives
+        , ("Word16be",      p prop_Word16be)
+        , ("Word16le",      p prop_Word16le)
+        , ("Word16host",    p prop_Word16host)
+        , ("Word32be",      p prop_Word32be)
+        , ("Word32le",      p prop_Word32le)
+        , ("Word32host",    p prop_Word32host)
+        , ("Word64be",      p prop_Word64be)
+        , ("Word64le",      p prop_Word64le)
+        , ("Word64host",    p prop_Word64host)
+        , ("Wordhost",      p prop_Wordhost)
+
+-- higher level ones using the Binary class
+        ,("()",         p (test :: T ()                     ))
+        ,("Bool",       p (test :: T Bool                   ))
+        ,("Ordering",   p (test :: T Ordering               ))
+
+        ,("Word8",      p (test :: T Word8                  ))
+        ,("Word16",     p (test :: T Word16                 ))
+        ,("Word32",     p (test :: T Word32                 ))
+        ,("Word64",     p (test :: T Word64                 ))
+
+        ,("Int8",       p (test :: T Int8                   ))
+        ,("Int16",      p (test :: T Int16                  ))
+        ,("Int32",      p (test :: T Int32                  ))
+        ,("Int64",      p (test :: T Int64                  ))
+
+        ,("Word",       p (test :: T Word                   ))
+        ,("Int",        p (test :: T Int                    ))
+        ,("Integer",    p (test :: T Integer                ))
+
+        ,("Float",      p (test :: T Float                  ))
+        ,("Double",     p (test :: T Double                 ))
+
+        ,("Char",       p (test :: T Char                   ))
+
+        ,("[()]",       p (test :: T [()]                  ))
+        ,("[Word8]",    p (test :: T [Word8]               ))
+        ,("[Word32]",   p (test :: T [Word32]              ))
+        ,("[Word64]",   p (test :: T [Word64]              ))
+        ,("[Word]",     p (test :: T [Word]                ))
+        ,("[Int]",      p (test :: T [Int]                 ))
+        ,("[Integer]",  p (test :: T [Integer]             ))
+        ,("String",     p (test :: T String                ))
+
+        ,("((), ())",           p (test :: T ((), ())        ))
+        ,("(Word8, Word32)",    p (test :: T (Word8, Word32) ))
+        ,("(Int8, Int32)",      p (test :: T (Int8,  Int32)  ))
+        ,("(Int32, [Int])",     p (test :: T (Int32, [Int])  ))
+
+        ,("Maybe Int8",         p (test :: T (Maybe Int8)        ))
+        ,("Either Int8 Int16",  p (test :: T (Either Int8 Int16) ))
+
+        ,("(Maybe Word8, Bool, [Int], Either Bool Word8)",
+                p (test :: T (Maybe Word8, Bool, [Int], Either Bool Word8) ))
+
+        ,("(Int, ByteString)",        p (test     :: T (Int, B.ByteString)   ))
+--      ,("Lazy (Int, ByteString)",   p (lazyTrip :: T (Int, B.ByteString)   ))
+        ,("[(Int, ByteString)]",      p (test     :: T [(Int, B.ByteString)] ))
+--      ,("Lazy [(Int, ByteString)]", p (lazyTrip :: T [(Int, B.ByteString)] ))
+
+
+--      ,("Lazy IntMap",       p (lazyTrip  :: T IntSet.IntSet          ))
+        ,("IntSet",            p (test      :: T IntSet.IntSet          ))
+        ,("IntMap ByteString", p (test      :: T (IntMap.IntMap B.ByteString) ))
+
+        ,("B.ByteString",  p (test :: T B.ByteString        ))
+        ,("L.ByteString",  p (test :: T L.ByteString        ))
+
+        ,("B.ByteString invariant",   p (prop_invariant :: B B.ByteString                 ))
+        ,("[B.ByteString] invariant", p (prop_invariant :: B [B.ByteString]               ))
+        ,("L.ByteString invariant",   p (prop_invariant :: B L.ByteString                 ))
+        ,("[L.ByteString] invariant", p (prop_invariant :: B [L.ByteString]               ))
+        ,("IntMap invariant",         p (prop_invariant :: B (IntMap.IntMap B.ByteString) ))
+
+        ,("Set Word32",      p (test :: T (Set.Set Word32)      ))
+        ,("Map Word16 Int",  p (test :: T (Map.Map Word16 Int)  ))
+
+        ,("(Maybe Int64, Bool, [Int])", p (test :: T (Maybe Int64, Bool, [Int])))
+
+{-
+--
+-- Big tuples lack an Arbitrary instance in Hugs/QuickCheck
+--
+
+        ,("(Maybe Word16, Bool, [Int], Either Bool Word16, Int)",
+            p (test :: T (Maybe Word16, Bool, [Int], Either Bool Word16, Int) ))
+
+        ,("(Maybe Word32, Bool, [Int], Either Bool Word32, Int, Int)", p (roundTrip :: (Maybe Word32, Bool, [Int], Either Bool Word32, Int, Int) -> Bool))
+
+        ,("(Maybe Word64, Bool, [Int], Either Bool Word64, Int, Int, Int)", p (roundTrip :: (Maybe Word64, Bool, [Int], Either Bool Word64, Int, Int, Int) -> Bool))
+-}
+
+-- GHC only:
+--      ,("Sequence", p (roundTrip :: Seq.Seq Int64 -> Bool))
+
+-- Obsolete
+--      ,("ensureLeft/Fail", mytest (shouldFail (decode L.empty :: Either ParseError Int)))
+        ]
diff --git a/tests/QuickCheckUtils.hs b/tests/QuickCheckUtils.hs
new file mode 100644
--- /dev/null
+++ b/tests/QuickCheckUtils.hs
@@ -0,0 +1,258 @@
+{-# OPTIONS_GHC -fglasgow-exts #-}
+--
+-- Uses multi-param type classes
+--
+module QuickCheckUtils where
+
+import Control.Monad
+
+import Test.QuickCheck.Batch
+import Test.QuickCheck
+import Text.Show.Functions
+
+import qualified Data.ByteString as B
+import qualified Data.ByteString.Unsafe as B
+import qualified Data.ByteString.Internal as B
+import qualified Data.ByteString.Lazy as L
+import qualified Data.Map as Map
+import qualified Data.Set as Set
+import qualified Data.IntMap as IntMap
+import qualified Data.IntSet as IntSet
+
+import qualified Control.Exception as C (evaluate)
+
+import Control.Monad        ( liftM2 )
+import Data.Char
+import Data.List
+import Data.Word
+import Data.Int
+import System.Random
+import System.IO
+
+-- import Control.Concurrent
+import System.Mem
+import System.CPUTime
+import Text.Printf
+
+import qualified Data.ByteString      as P
+import qualified Data.ByteString.Lazy as L
+import qualified Data.ByteString.Lazy.Internal as L
+
+-- import qualified Data.Sequence as Seq
+
+-- Enable this to get verbose test output. Including the actual tests.
+debug = False
+
+mytest :: Testable a => a -> Int -> IO ()
+mytest a n = mycheck defaultConfig
+    { configMaxTest=n
+    , configEvery= \n args -> if debug then show n ++ ":\n" ++ unlines args else [] } a
+
+mycheck :: Testable a => Config -> a -> IO ()
+mycheck config a = do
+     rnd <- newStdGen
+     performGC -- >> threadDelay 100
+     t <- mytests config (evaluate a) rnd 0 0 [] 0 -- 0
+     printf " %0.3f seconds\n" (t :: Double)
+     hFlush stdout
+
+time :: a -> IO (a , Double)
+time a = do
+    start <- getCPUTime
+    v     <- C.evaluate a
+    v `seq` return ()
+    end   <- getCPUTime
+    return (v,     (      (fromIntegral (end - start)) / (10^12)))
+
+mytests :: Config -> Gen Result -> StdGen -> Int -> Int -> [[String]] -> Double -> IO  Double
+mytests config gen rnd0 ntest nfail stamps t0
+  | ntest == configMaxTest config = do done "OK," ntest stamps
+                                       return t0
+
+  | nfail == configMaxFail config = do done "Arguments exhausted after" ntest stamps
+                                       return t0
+
+  | otherwise = do
+     (result,t1) <- time (generate (configSize config ntest) rnd2 gen)
+
+     putStr (configEvery config ntest (arguments result)) >> hFlush stdout
+     case ok result of
+       Nothing    ->
+         mytests config gen rnd1 ntest (nfail+1) stamps (t0 + t1)
+       Just True  ->
+         mytests config gen rnd1 (ntest+1) nfail (stamp result:stamps) (t0 + t1)
+       Just False -> do
+         putStr ( "Falsifiable after "
+               ++ show ntest
+               ++ " tests:\n"
+               ++ unlines (arguments result)
+                ) >> hFlush stdout
+         return t0
+
+     where
+      (rnd1,rnd2) = split rnd0
+
+done :: String -> Int -> [[String]] -> IO ()
+done mesg ntest stamps = putStr ( mesg ++ " " ++ show ntest ++ " tests" ++ table )
+ where
+  table = display
+        . map entry
+        . reverse
+        . sort
+        . map pairLength
+        . group
+        . sort
+        . filter (not . null)
+        $ stamps
+
+  display []  = ". "
+  display [x] = " (" ++ x ++ "). "
+  display xs  = ".\n" ++ unlines (map (++ ".") xs)
+
+  pairLength xss@(xs:_) = (length xss, xs)
+  entry (n, xs)         = percentage n ntest
+                       ++ " "
+                       ++ concat (intersperse ", " xs)
+
+  percentage n m        = show ((100 * n) `div` m) ++ "%"
+
+------------------------------------------------------------------------
+
+instance Random Word8 where
+  randomR = integralRandomR
+  random = randomR (minBound,maxBound)
+
+instance Random Int8 where
+  randomR = integralRandomR
+  random = randomR (minBound,maxBound)
+
+instance Random Word16 where
+  randomR = integralRandomR
+  random = randomR (minBound,maxBound)
+
+instance Random Int16 where
+  randomR = integralRandomR
+  random = randomR (minBound,maxBound)
+
+instance Random Word where
+  randomR = integralRandomR
+  random = randomR (minBound,maxBound)
+
+instance Random Word32 where
+  randomR = integralRandomR
+  random = randomR (minBound,maxBound)
+
+instance Random Int32 where
+  randomR = integralRandomR
+  random = randomR (minBound,maxBound)
+
+instance Random Word64 where
+  randomR = integralRandomR
+  random = randomR (minBound,maxBound)
+
+instance Random Int64 where
+  randomR = integralRandomR
+  random = randomR (minBound,maxBound)
+
+------------------------------------------------------------------------
+
+integralRandomR :: (Integral a, RandomGen g) => (a,a) -> g -> (a,g)
+integralRandomR  (a,b) g = case randomR (fromIntegral a :: Integer,
+                                         fromIntegral b :: Integer) g of
+                            (x,g) -> (fromIntegral x, g)
+
+------------------------------------------------------------------------
+
+instance Arbitrary Word8 where
+    arbitrary       = choose (0, 2^8-1)
+    coarbitrary w   = variant 0
+
+instance Arbitrary Word16 where
+    arbitrary       = choose (0, 2^16-1)
+    coarbitrary     = undefined
+
+instance Arbitrary Word32 where
+--  arbitrary       = choose (0, 2^32-1)
+    arbitrary       = choose (minBound, maxBound)
+    coarbitrary     = undefined
+
+instance Arbitrary Word64 where
+--  arbitrary       = choose (0, 2^64-1)
+    arbitrary       = choose (minBound, maxBound)
+    coarbitrary     = undefined
+
+instance Arbitrary Int8 where
+--  arbitrary       = choose (0, 2^8-1)
+    arbitrary       = choose (minBound, maxBound)
+    coarbitrary w   = variant 0
+
+instance Arbitrary Int16 where
+--  arbitrary       = choose (0, 2^16-1)
+    arbitrary       = choose (minBound, maxBound)
+    coarbitrary     = undefined
+
+instance Arbitrary Int32 where
+--  arbitrary       = choose (0, 2^32-1)
+    arbitrary       = choose (minBound, maxBound)
+    coarbitrary     = undefined
+
+instance Arbitrary Int64 where
+--  arbitrary       = choose (0, 2^64-1)
+    arbitrary       = choose (minBound, maxBound)
+    coarbitrary     = undefined
+
+instance Arbitrary Word where
+    arbitrary       = choose (minBound, maxBound)
+    coarbitrary w   = variant 0
+
+------------------------------------------------------------------------
+
+instance Arbitrary Char where
+    arbitrary = choose (maxBound, minBound)
+    coarbitrary = undefined
+
+{-
+instance Arbitrary a => Arbitrary (Maybe a) where
+    arbitrary = oneof [ return Nothing, liftM Just arbitrary]
+    coarbitrary = undefined
+    -}
+
+instance Arbitrary Ordering where
+    arbitrary = oneof [ return LT,return  GT,return  EQ ]
+    coarbitrary = undefined
+
+{-
+instance (Arbitrary a, Arbitrary b) => Arbitrary (Either a b) where
+    arbitrary = oneof [ liftM Left arbitrary, liftM Right arbitrary]
+    coarbitrary = undefined
+    -}
+
+instance Arbitrary IntSet.IntSet where
+    arbitrary = fmap IntSet.fromList arbitrary
+    coarbitrary = undefined
+
+instance (Arbitrary e) => Arbitrary (IntMap.IntMap e) where
+    arbitrary = fmap IntMap.fromList arbitrary
+    coarbitrary = undefined
+
+instance (Arbitrary a, Ord a) => Arbitrary (Set.Set a) where
+    arbitrary = fmap Set.fromList arbitrary
+    coarbitrary = undefined
+
+instance (Arbitrary a, Ord a, Arbitrary b) => Arbitrary (Map.Map a b) where
+    arbitrary = fmap Map.fromList arbitrary
+    coarbitrary = undefined
+
+{-
+instance (Arbitrary a) => Arbitrary (Seq.Seq a) where
+    arbitrary = fmap Seq.fromList arbitrary
+    coarbitrary = undefined
+-}
+
+instance Arbitrary L.ByteString where
+    arbitrary     = arbitrary >>= return . L.fromChunks . filter (not. B.null) -- maintain the invariant.
+    coarbitrary s = coarbitrary (L.unpack s)
+
+instance Arbitrary B.ByteString where
+  arbitrary = B.pack `fmap` arbitrary
+  coarbitrary s = coarbitrary (B.unpack s)
diff --git a/todo b/todo
new file mode 100644
--- /dev/null
+++ b/todo
@@ -0,0 +1,28 @@
+layer handling:
+
+    bit packing
+    state parameters
+    string pools
+
+    reading structures from the end of a stream, seek/tell behaviour
+
+seek based protocols are too hard. 
+    hGetContents/ interleaving.
+
+user requests:
+
+    get remaining bytestring after a runGet
+
+    some kind of lookahead, or restoring parsing state, or something with
+      equal functionality. make it another layer on top?
+
+    getLazyByteString takes an Int, which in Haskell98 is only guarantied to
+      be 29 bits, ie. 512 mb.
+      maybe we should have a readN64 for allowing reading of larger stuff?
+      (which could be implemented with readN on 64bit machines)
+      reference: bringerts tar archive decoder would be limitid to 0.5GB
+                 files, alt. 2GB in GHC
+
+SYB-deriving
+
+investigate the UArray instance, it does not seem to compile in GHC 6.4
diff --git a/tools/derive/BinaryDerive.hs b/tools/derive/BinaryDerive.hs
new file mode 100644
--- /dev/null
+++ b/tools/derive/BinaryDerive.hs
@@ -0,0 +1,57 @@
+{-# OPTIONS -fglasgow-exts #-}
+
+module BinaryDerive where
+
+import Data.Generics
+import Data.List
+
+deriveM ::  (Typeable a, Data a) => a -> IO ()
+deriveM (a :: a) = mapM_ putStrLn . lines $ derive (undefined :: a)
+
+derive :: (Typeable a, Data a) => a -> String
+derive x = 
+    "instance " ++ context ++ "Binary " ++ inst ++ " where\n" ++
+    concat putDefs ++ getDefs
+    where
+    context
+        | nTypeChildren > 0 =
+            wrap (join ", " (map ("Binary "++) typeLetters)) ++ " => "
+        | otherwise = ""
+    inst = wrap $ tyConString typeName ++ concatMap (" "++) typeLetters
+    wrap x = if nTypeChildren > 0 then "("++x++")" else x 
+    join sep lst = concat $ intersperse sep lst
+    nTypeChildren = length typeChildren
+    typeLetters = take nTypeChildren manyLetters
+    manyLetters = map (:[]) ['a'..'z']
+    (typeName,typeChildren) = splitTyConApp (typeOf x)
+    constrs :: [(Int, (String, Int))]
+    constrs = zip [0..] $ map gen $ dataTypeConstrs (dataTypeOf x)
+    gen con = ( showConstr con
+              , length $ gmapQ undefined $ fromConstr con `asTypeOf` x
+              )
+    putDefs = map ((++"\n") . putDef) constrs
+    putDef (n, (name, ps)) =
+        let wrap = if ps /= 0 then ("("++) . (++")") else id
+            pattern = name ++ concatMap (' ':) (take ps manyLetters)
+        in
+        "  put " ++ wrap pattern ++" = "
+        ++ concat [ "putWord8 " ++ show n | length constrs  > 1 ]
+        ++ concat [ " >> "                | length constrs  > 1 && ps  > 0 ]
+        ++ concat [ "return ()"           | length constrs == 1 && ps == 0 ]
+        ++ join " >> " (map ("put "++) (take ps manyLetters))
+    getDefs =
+       (if length constrs > 1
+            then "  get = do\n    tag_ <- getWord8\n    case tag_ of\n"
+            else "  get =")
+        ++ concatMap ((++"\n")) (map getDef constrs) ++
+       (if length constrs > 1
+	    then "      _ -> fail \"no parse\""
+	    else ""
+       )
+    getDef (n, (name, ps)) =
+        let wrap = if ps /= 0 then ("("++) . (++")") else id
+        in
+        concat [ "      " ++ show n ++ " ->" | length constrs > 1 ]
+        ++ concatMap (\x -> " get >>= \\"++x++" ->") (take ps manyLetters)
+        ++ " return "
+        ++ wrap (name ++ concatMap (" "++) (take ps manyLetters))
diff --git a/tools/derive/Example.hs b/tools/derive/Example.hs
new file mode 100644
--- /dev/null
+++ b/tools/derive/Example.hs
@@ -0,0 +1,68 @@
+
+import Data.Generics
+
+import Data.Binary
+
+import BinaryDerive
+
+data Foo = Bar
+    deriving (Typeable, Data, Show, Eq)
+
+instance Binary Main.Foo where
+  put Bar = return ()
+  get = return Bar
+
+data Color = RGB Int Int Int
+           | CMYK Int Int Int Int
+    deriving (Typeable, Data, Show, Eq)
+
+instance Binary Main.Color where
+  put (RGB a b c) = putWord8 0 >> put a >> put b >> put c
+  put (CMYK a b c d) = putWord8 1 >> put a >> put b >> put c >> put d
+  get = do
+    tag_ <- getWord8
+    case tag_ of
+      0 -> get >>= \a -> get >>= \b -> get >>= \c -> return (RGB a b c)
+      1 -> get >>= \a -> get >>= \b -> get >>= \c -> get >>= \d -> return (CMYK a b c d)
+
+data Computer = Laptop { weight :: Int }
+              | Desktop { speed :: Int, memory :: Int }
+    deriving (Typeable, Data, Show, Eq)
+
+instance Binary Main.Computer where
+  put (Laptop a) = putWord8 0 >> put a
+  put (Desktop a b) = putWord8 1 >> put a >> put b
+  get = do
+    tag_ <- getWord8
+    case tag_ of
+      0 -> get >>= \a -> return (Laptop a)
+      1 -> get >>= \a -> get >>= \b -> return (Desktop a b)
+
+-- | All drinks mankind will ever need
+data Drinks = Beer Bool{-ale?-}
+            | Coffee
+            | Tea
+            | EnergyDrink
+            | Water
+            | Wine
+            | Whisky
+    deriving (Typeable, Data, Show, Eq)
+
+instance Binary Main.Drinks where
+  put (Beer a) = putWord8 0 >> put a
+  put Coffee = putWord8 1
+  put Tea = putWord8 2
+  put EnergyDrink = putWord8 3
+  put Water = putWord8 4
+  put Wine = putWord8 5
+  put Whisky = putWord8 6
+  get = do
+    tag_ <- getWord8
+    case tag_ of
+      0 -> get >>= \a -> return (Beer a)
+      1 -> return Coffee
+      2 -> return Tea
+      3 -> return EnergyDrink
+      4 -> return Water
+      5 -> return Wine
+      6 -> return Whisky
