[Shootout-list] Directions of various benchmarks

[Einar Karttunen]

Jon Harrop <jon@ffconsultancy.com> writes:
> Which ones?

The list tests and hash2.

>> I would like to see 
>> more activity on e.g. graphs or trees with concurrent updates.
>
> Regarding trees, most languages (e.g. C, Fortran, C++, Java, C#) will not be 
> able to express tree-based programs in 100 LOC unless they use their standard 
> library, in which case it will be using specified functionality and you 
> probably can't guarantee its internal representation. I would not object to 
> some programs being represented in OCaml by a balanced binary tree and in 
> C/Fortran by a hash table, for example.

Well we cannot guarantee the internal representation of Strings either
and still we are using them ;) In some languages they are linked lists
not arrays... 

Sets and maps are quite common in real world problems and reflecting how
easily and efficiently they can be used in different languages would be
nice. Picking a suitable benchmark is harder...

> I've think we've decided that this isn't going to happen as the vast majority 
> of languages will not be able to express this in <100 LOC and most of the 
> remaining languages would use lex and yacc to get the job done.

I think it could be expressed even in C with <100 LOC using reverse
polish syntax. One could e.g. have fibonacci(some_suitable_number) as
input. I am not saying that we should have this as a test, just that
adding yet another number crunching test does not seem very good.

- Einar Karttunen