On Oct 30, 2008, at 5:21 PM, Bertram Felgenhauer wrote:
George Pollard wrote:
There's also the ieee-utils package, which provides an IEEE monad with `setRound`:
http://hackage.haskell.org/packages/archive/ieee-utils/0.4.0/doc/ html/Numeric-IEEE-RoundMode.html
When run with +RTS -N2 -RTS, the output randomly alternates between Downward and ToNearest - for me at least.
The problem is that the setRound call will only affect one worker thread, while the RTS will sometimes migrate RTS threads from one worker to another.
runIEEE really has to be executed in a bound thread (see forkOS documentation). Using `par` will also cause trouble - in fact even more.
That's a really nice catch! Dons has pointed out to me both the very handy forkOnIO which ensures the forked thread remains bound to a single CPU, and also the -qm flag to the RTS, which prevents thread migration between capabilities. Running the example program with +RTS -N2 -qm restores the behavior to what's intended. I'll try to get around to changing the documentation to reflect this. Also, it's worth noting that the IEEE round mode has no effect on rounding done with the `round` function, as that's explicitly coded to provide the behavior seen in the report. Regards, Sterl.