> socket Handles are unbuffered by the default, so
> if you change the buffering by inserting
>
> hSetBuffering handle
> (BlockBuffered Nothing) -- or some such
>
> after each call to socketToHandle / Socket.accept, you
> should get a boost in throughput.
Definitely - using the unbuffered sockets is likely to be slow due to the
overhead of GHC's I/O library.
If you're sending files over a socket, you can avoid the conversion to/from
Haskell Strings altogether, using GHC's hGetBuf/hPutBuf family (this is some
code from the web server):
bufsize = 4 * 1024 :: Int
-- squirt data from 'rd' into 'wr' as fast as possible. We use a 4k
-- single buffer.
squirt rd wr = do
arr <- stToIO (newCharArray (0, bufsize-1))
let loop = do r <- hGetBufBAFull rd arr bufsize
if (r == 0)
then return ()
else if (r < bufsize)
then hPutBufBAFull wr arr r
else hPutBufBAFull wr arr bufsize >> loop
loop
Cheers,
Simon
_______________________________________________
Glasgow-haskell-bugs mailing list
[EMAIL PROTECTED]
http://www.haskell.org/mailman/listinfo/glasgow-haskell-bugs
