I get no speedup with RWH examples of parallel sorting

I've carefully followed the instructions in Real World Haskell for parallel sorting: In SortMain.hs 1. Set testfunction=sort Then del *.o (force recompilation) ghc --make -O2 sortMain.hs run SortMain and I get 500000 numbers => 3.9735 sec. 2. Set testFunction = parSort2 2 del *.o ghc -threaded --make -O2 sortMain.hs run SortMain +RTS -N2 -RTS, I get => 5.3 sec run SortMain +RTS -N1 -RTS, I get => 4.3 sec With the +RTS -s option I get 3 SPARK3 (3 converted, 0 pruned) All other testFunctions are worse. I have a dual core computer that runs Windows XP. I've experimented a lot with par, pseq ... in other programs. I almost always get a few hundred sparks. Half of them are converted and the other half gets pruned. But I never got a faster program (then sequential). What am I doing wrong, any ideas? Kees

My experience with running parallel programs on dual processor machines is similar. With only two processors, one is always busy doing something for the OS. As a result, it rarely gets scheduled. You need a quad cpu to play with parallel programming. ____________________ David Place Owner, Panpipes Ho! LLC http://panpipesho.com d@vidplace.com On Jul 6, 2011, at 11:08 AM, Kees Bleijenberg wrote:
I've carefully followed the instructions in Real World Haskell for parallel sorting: In SortMain.hs 1. Set testfunction=sort Then del *.o (force recompilation) ghc --make -O2 sortMain.hs run SortMain and I get 500000 numbers => 3.9735 sec.
2. Set testFunction = parSort2 2 del *.o ghc -threaded --make -O2 sortMain.hs run SortMain +RTS -N2 -RTS, I get => 5.3 sec run SortMain +RTS -N1 -RTS, I get => 4.3 sec With the +RTS -s option I get 3 SPARK3 (3 converted, 0 pruned) All other testFunctions are worse.
I have a dual core computer that runs Windows XP. I've experimented a lot with par, pseq ... in other programs. I almost always get a few hundred sparks. Half of them are converted and the other half gets pruned. But I never got a faster program (then sequential).
What am I doing wrong, any ideas?
Kees
_______________________________________________ Beginners mailing list Beginners@haskell.org http://www.haskell.org/mailman/listinfo/beginners
participants (2)
-
David Place
-
Kees Bleijenberg