[Haskell-cafe] IORef vs TVar performance: 6 seconds versus 4 minutes

Evan Laforge qdunkan at gmail.com
Mon Dec 29 23:28:40 EST 2008


> In the short term, yes.  But it's my opinion that most anything that's
> a win for ease of modification is a win for performance in the long
> run, if it gets more people writing code.

Also agreed, in theory.  The language is definitely also in line with
this philosophy.  In practice, though, it's also nice for low level
stuff to be fast right now.  However, I expect that if the haskell RTS
gets implemented it'll be a switch at least until it's as fast or
faster than the C version, so it's all good.

Anyway, thanks for the clarification on the STM performance.  I guess
it means I can't brag too much about pure performance, but there's
plenty else to brag about.


More information about the Haskell-Cafe mailing list