> Could someone please tell me the status of the various
> parellel modes of
> ghc 4.08? I remember building the smp version of 4.07 a while
> back, and
> while I could build a gph program under the smp-enable
> compiler, and I did
> observe context-switches occurring more often, it did not seem to take
> advantage of the native (linux) threads.
The SMP support in GHC is still work-in-progress, I hope to get back to it
sometime this year. It still has a number of problems related to
non-blocking IO and signal delivery, the performance needs a lot of tuning,
and there are still a few races and deadlocks lurking. You might be able to
get it working on simple examples that don't do much I/O, though. To enable
threads, you need to run the program with +RTS -N2 (for 2 threads), and
preferably -Sstderr to get timing stats for each thread.
GPH in GHC 4.xx is also work-in-progress; Hans Lloidl is working on
integrating the old 3.xx code into the new tree. The GPH folks have their
own mailing list: [EMAIL PROTECTED] (use [EMAIL PROTECTED] to
subscribe).
Cheers,
Simon
