[erlang-questions] Erlang math libraries

jm jeffm@REDACTED
Thu May 17 02:46:29 CEST 2007


ok wrote:
> Mind you, it would be nice to have good numeric performance *as well*,
> and one could certainly imagine an amalgam of Erlang and APL (but not,
> please, J; there are good ideas in J but readability isn't one of them).
> 
> Let's take an example.
> 
> 8000 dot products of 3000-element vectors:
>     Fortran 90:  0.79 seconds  (Sun Fortran compiler, -O3)
>     C 89:        1.96 seconds  (Sun C compiler, -xO3)
>     Erlang:     42.33 seconds  (erlc +native)
> 
> You would have to be running on at least 54 machines (assuming linear
> speedup) for parallel Erlang code to catch up with Fortran, but then
> the Fortran compiler I use would do automatic parallelisation for me
> if I had the right licence (quick check: by golly I *DO*, it's just
> that I only have a uniprocessor machine).

That is simple demostration of different languages are better suited, at
least todays laguages, to different problem domains. Erlang while
developed it a telco environment I actually think of as a control
systems langauge for conplex systems. So in is no surprise to be to see
it being pick up by web developers. What does suprise me though is that
I haven't seen many references to robotics. May be it's just that the
web guys make more noise on the mailing list :-).

Turning back to the topic at hand. It would be interesting to see what a
grid computing scenario with erlang with fortran libraries would be
like. Erlang could do the monitoring and control, passing parameters
around, etcetera while Fortran with it's execellent maths libraries
would do the grunt work.


Jeff.



More information about the erlang-questions mailing list