Re: [Gretl-devel] Collection of gretl speed examples
Am 08.09.2017 um 14:31 schrieb Riccardo (Jack) Lucchetti:
It'd be also interesting to see what happens using
Matlab instead of Octave (if you're not ideologically averse to the use
of non-free software).
True; however, once you have a ratio for gretl/Octave for a certain
task, I think for the Matlab people it would be fairly easy to infer the
gretl/Matlab ratio by calculating gretl/Octave * Octave/Matlab. So I
don't view this as essential, at least initially.
Anyway, my script is just a little toy thingie that I use to support
the
claim that hansl is a perfectly good language for serious number
crunching.
Don't know why you are dismissing this as a "toy". It is very
interesting and useful, and also shows how gretl can glue together other
languages via 'foreign'.
However, there are so many more things that could (and
should) be tested, in order to have a proper comparison.
"could" be tested, yes; but "should" be tested, I don't agree. My
idea
was to leverage what people have anyway. The aim is a low-effort
collective action, otherwise I believe this will never take off, because
who has the time to work on a "proper" comparison. As you use to say,
release early and release often, so we can fill in any gaps later.
http://www.cybertester.com/data/ncrunch4.pdf
It'd be nice to use some of that too.
Well, it's very old. I don't want to stop people from sending
comparisons with Gauss, but I certainly am not going to work on that
myself. (Admittedly, there's some Ox and Matlab stuff as well.)
thanks,
sven