[Shootout-list] Directions of various benchmarks
Jon Harrop
jon@ffconsultancy.com
Mon, 16 May 2005 01:51:37 +0100
On Sunday 15 May 2005 18:08, Isaac Gouy wrote:
> --- Jon Harrop <jon@ffconsultancy.com> wrote:
> > On Sunday 15 May 2005 09:16, Einar Karttunen wrote:
> > > and most of the new ones seem to be array related number crunching.
> >
> > Yes, although the ray tracer is FP-bound but is based on a scene
> > tree.
>
> No.
>
> fasta (new)
Most of the time is spent in the random number generator = number crunching.
> k-nucleotide (new)
Most of the time is spent hashing strings = array-based number crunching.
> pidigits
Arbitrary-precision integers = arrays
arithmetic over them = number crunching
> reverse-complement (new)
If this didn't take an immeasurably small time to run then it would be IO
bound.
> tcp-echo (new)
> tcp-request-reply (new)
> tcp-stream (new)
> threads
> threads-flow
Yes, I always disregard these benchmarks.
You also missed implicitode. That is obviously number crunching.
To the best of my knowledge the informal <100LOC of C#/Java rule has
prohibited the use of non-trivial data structures in all of the benchmarks as
they are intractable to implement in those languages. My interpretation of
the OP is that he would like to see more benchmarks which spend their time
rebalancing trees. I agree. The OCaml implementation of my "nth" benchmark
will do this.
--
Dr Jon D Harrop, Flying Frog Consultancy Ltd.
Objective CAML for Scientists
http://www.ffconsultancy.com/products/ocaml_for_scientists