[Haskell-beginners] Performance of parallel mergesort

[Jon Harrop]

On Thursday 24 December 2009 09:09:41 Tom Davie wrote:
> Here's my results on a dual core running OS X:
>
> LappyBob:~ tatd2$ time ./test +RTS -N1
> -0.117109518058233
>
> real 0m4.608s
> user 0m4.400s
> sys 0m0.189s
> LappyBob:~ tatd2$ time ./test +RTS -N2
> -0.117109518058233
>
> real 0m3.648s
> user 0m6.360s
> sys 0m0.220s
> LappyBob:~ tatd2$ time ./test +RTS -N3
> -0.117109518058233
>
> real 0m50.679s
> user 1m24.235s
> sys 0m0.620s

Here are my results on a dual quad-core:

$ ghc-6.10.4 --make -O2 -threaded mergesort.hs -o mergesort
[1 of 1] Compiling Main             ( mergesort.hs, mergesort.o )
Linking mergesort ...
jdh30 at leper:~/Programs/mine/haskell/mergesort$ time ./mergesort 
+RTS -K100000000 +RTS -N1
-0.117109518058233

real    0m7.664s
user    0m7.528s
sys     0m0.136s
$ time ./mergesort +RTS -K100000000 +RTS -N2
-0.117109518058233

real    0m7.224s
user    0m9.773s
sys     0m0.108s
$ time ./mergesort +RTS -K100000000 +RTS -N3
-0.117109518058233

real    0m7.293s
user    0m11.637s
sys     0m0.168s
$ time ./mergesort +RTS -K100000000 +RTS -N4
-0.117109518058233

real    0m7.333s
user    0m12.689s
sys     0m0.360s
$ time ./mergesort +RTS -K100000000 +RTS -N5
-0.117109518058233

real    0m7.449s
user    0m14.757s
sys     0m0.548s
$ time ./mergesort +RTS -K100000000 +RTS -N6
-0.117109518058233

real    0m7.043s
user    0m14.985s
sys     0m0.252s
$ time ./mergesort +RTS -K100000000 +RTS -N7
-0.117109518058233

real    0m8.018s
user    0m19.325s
sys     0m1.012s
$ time ./mergesort +RTS -K100000000 +RTS -N8
-0.117109518058233

real    0m55.963s
user    5m55.654s
sys     0m1.492s

As you can see, there is still no significant speedup from parallelism.

-- 
Dr Jon Harrop, Flying Frog Consultancy Ltd.
http://www.ffconsultancy.com/?e