[Shootout-list] some perl test

[William Douglas Neumann]

On Mon, 29 Nov 2004, Isaac Gouy wrote:

> If we were doing Math instead of timing recursive function calls - we
> use ints for Ackermann and Fibonacci

Then perhaps you should allow the implemetor to choose the datatype, 
because as it stands now, you are not timing "recursive function calls". 
For example, in OCaml if you are using floats, you are testing both the 
timing of recursive function calls *and* the timing of unboxing floats. 
Running takfp takes over 2.3 times longer under OCaml than tak_int.  So 
you should either allow for different data types or change the description 
to reflect that you are also testing floating point performance.

> "Historical note: <snip>

Then it might be best to rewrite the "about the benchmark" to explictly 
state that the TAK function at the Wolfram site is not what is being 
tested, but rather a variant of that function.  I looked at those notes 
and thought, "Well, I couldn't give a rat's ass about the origins of the 
TAK function, so I'll go to Wolfram to see what it is."  So I went, saw, 
coded it up and then looked at the Java code where I saw that passing in h 
was not required.  I'd assume others are likely to do the same.


William D. Neumann
<wdnx@unm.edu>

---
Dear Lord, please make me the kind of person
my dog thinks I am.