Re: pidigits benchmark

["Mark Meijer" <meijer78@...>]

I see your point. But then you're talking about benchmarking some
library and/or specific bindings for it to other programming
languages. Useful info, to be sure, but not the same as benchmarking
those programming languages themselves. All languages can be blazingly
fast if everything you need from it is implemented in native
libraries. If that is the point of some benchmarking site, no problem.
All I'm saying is there's a clear distinction that tends to be
overlooked.

If I needed library X for some project, and I want to find out how
different bindings to that library perform in different languages,
then I'm interested in benchmarking those. If I want to compare
benchmarks of different programming languages in general (if there is
such a thing - of course there is always a context for one's interest
in comparing languages, that's basically why benchmarks measure
different algorithms), then I'm not interested in seeing benchmarks
that largely measure the performance of some native libraries. I'd
much rather see those algorithms implemented in pure <insert language>
(e.g. pure lua, pure python, etc) for the purpose of benchmarking
those languages. To me, that's a big difference.

I'm assuming that's also what Mike Pall meant when he said "This
benchmark is almost meaningless for ***comparing languages*** because
it spends most of the time in GMP" (emphasis mine).


2008/8/25 Isaac Gouy <igouy2@yahoo.com>:
> 2008/8/18 Mark Meijer:
>
>> Indeed, what is the point of benchmarking a scripting engine using
>> algorithms implemented in "native" libraries.
>
>> Seems kinda silly to me.
>
> Although I have some sympathy with the view that language benchmarking
> is silly, period; this kind of complaint is still a little
> disappointing because it's so unimaginative.
>
> The point is:
> - to see how much difference there is between Lua+GMP and Lua
> - to see how much difference there is between Lua and gcc+GMP
> - to see how much less difference there is between Lua+GMP and gcc+GMP
> - to see if the programs blow-up (GNU Smalltalk memory use)
> - to see what's going on with the different programs
> etc etc
>
>
> 2008/8/18 Mike Pall <mikelu-0808@mike.de>:
>
>> This benchmark is almost meaningless for comparing languages
>> because it spends most of the time in GMP (for those programs that
>> use it, i.e. almost all of them).
>
> You seem to be suggesting that it's meaningless to show there's not a
> massive difference between gcc+GMP and Lua+GMP - I kind-of thought that
> was the point of glue languages ;-)
>
>
>
>
>