Haskell performance

Malcolm Wallace Malcolm.Wallace at cs.york.ac.uk
Thu Mar 18 11:51:15 EST 2004

Josef Svenningsson <josefs at cs.chalmers.se> writes:

> > > [Doug Bagley's Language Shootout]
> >
> > You should look at the individual examples and see how relevant their
> > results are for you.
> Well, I think this shows that one should be very careful when reading
> these kinds of benchmarks.

And don't forget that the performance results are probably quite
out-of-date now in any case.  I took a quick look at just one benchmark
from the Shootout: Nested Loops.  The CPU times given are:

    gcc  0.10
    ghc  0.34

Directly taking the given source code and compiling it afresh today,
with -O turned on for both compilers, like in the original tests,
I get results like these:

    gcc  0.90
    ghc  1.70

Since I'm using a faster machine, I needed to increase the input size
from N=16 to N=30 to get measurable timings, but the ratio between
ghc and ghc today is now consistently about 2x for this benchmark,
a big improvement over 3.4x.  I guess the ghc implementers are to be


More information about the Glasgow-haskell-users mailing list