[Shootout-list] Stuff

[Brent Fulgham]

--- Jon Harrop <jon@ffconsultancy.com> wrote:

> For me, these "same thing" benchmarks are more
> interesting than the "same way" 
> benchmarks because I'd attack problems using
> different techniques depending 
> upon the language I was writing in.

Yes.  I think we should try to minimize the number of
'same way' tests in the shootout for this reason.

There are obvious cases where 'same way' is needed,
since we are trying to show the performance of some
facet of a language (e.g., method invocation, or
recursive calling, or string concatenation, etc.)

I would like to see us incorporate a few 'larger'
benchmarks that are 'same thing' tests that allow
each implementation to do things the 'natural' way.

We have been trying to get rid of the various tests
that required looping over the same data multiple
times to measure performance, because good compilers
see this and remove the redundant work (e.g.,
Haskell is a terror for us since it does this so
well).

I think the larger problems proposed by Sebastien
and Jon are good steps forward to performing these
types of measurements.

-Brent