Re: [Haskell-cafe] Re: Thread pool in GHC

6 Sep 2005


      Yes, but AFAIK, threads in Haskell are exceedingly lightweight, as
they are only primitives for concurrency. To be perfectly honest, I've
found that my solution is completely useless in practise. I made up
the solution cos I reconned that tweaking GHC runtime heap size would
be harder than making a threadpool (it isn't). Basically, all the
usual reasons for using a threadpool just don't apply in Haskell.

Gen

On 9/6/05, Dinh Tien Tuan Anh  wrote:
...
Does it mean that the thread pool will "LAUNCH" new thread as long as the
thread count does not exceed a maximum number ?
If it does launch the new thread, then the idea of thread reuse is
completely ignored.
My program creates and kills at most 5 threads at a time, but this process
repeated for an infinite times, so thread reuse would improve the
performance by getting of the overhead of creating new thread.
TuanAnh
...
From: genneth 
To: Dinh Tien Tuan Anh 
CC: haskell-cafe@haskell.org
Subject: Re: [Haskell-cafe] Re: Thread pool in GHC
Date: Tue, 6 Sep 2005 10:25:15 +0800
I think it would go back to the pool. The finally clause means that
even if the thread dies from an exception (which, AFAIK, is how kills
are modelled), the thread count would be restored.
Gen
On 9/6/05, Dinh Tien Tuan Anh  wrote:
...
Its probably too long to bring back this topic, but i have a small
question.
...
If some threads may never terminate and have to be killed by killThread,
are
...
they going back to the pool, or we need some twist to force them.
Thanks a lot
TuanAnh
...
From: genneth 
To: haskell-cafe@haskell.org
Subject: [Haskell-cafe] Re: Thread pool in GHC
Date: Thu, 4 Aug 2005 16:47:56 +0000 (UTC)
Dinh Tien Tuan Anh  writes:
...
Can thread pool be implemented in GHC ?
I have a program that is currently using about 12-15 threads (launch
and
...
kill for infinite times) and when running, especially after Ctrl-C,
my
computer got freezed up. And if i ran it several times, the "Stack
overflows" occurs.
I made the following a while back. Maybe it's useful...
limitedThreadsWithChannelMapM :: Integer -> (a -> IO b) -> [a] -> IO
[MVar
b]
limitedThreadsWithChannelMapM lim ioaction x = do
    threadpoolcounter <- atomically ( newTVar 0 )
    mapM (throttledFork threadpoolcounter . ioaction) x
    where
        throttledFork poolcount io = do
            atomically ( do
                prev <- readTVar poolcount
                if prev >= lim then
                    retry
                    else writeTVar poolcount (prev+1) )
            mvar <- newEmptyMVar
            forkIO(
                finally
                    (io >>= putMVar mvar)
                    (atomically ( readTVar poolcount >>= writeTVar
poolcount .
(subtract 1) ) ) )
            return mvar
...
Cheers
TuanAnh
_________________________________________________________________
Winks & nudges are here - download MSN Messenger 7.0 today!
http://messenger.msn.co.uk
_______________________________________________
Haskell-Cafe mailing list
Haskell-Cafe@haskell.org
http://www.haskell.org/mailman/listinfo/haskell-cafe
_________________________________________________________________
Winks & nudges are here - download MSN Messenger 7.0 today!
http://messenger.msn.co.uk
_________________________________________________________________
It's fast, it's easy and it's free. Get MSN Messenger 7.0 today!
http://messenger.msn.co.uk