[Haskell-cafe] Re: Processing of large files

[Simon Marlow]

John Goerzen writes:
 
> Yes it does.  If you don't set block buffering, GHC will call 
> read() separately for *every* single character.  (I've 
> straced stuff!)  This is a huge performance penalty for large 
> files.  It's a lot more efficient if you set block buffering 
> in your input, even if you are using interact and lines or 
> words to process it.

That shouldn't be the case.  GHC's IO library checks whether a Handle
refers to a TTY, and enables line buffering if it does.  Otherwise,
block buffering is the default.

Could you isolate the test case and submit a bug report?

Cheers,
	Simon