[Shootout-list] fannkuch benchmark for GHC

[Greg Buchholz]

--- Isaac Gouy <igouy2@yahoo.com> wrote:
> Fannkuch is supposed to be a test of indexed-access to a collection.

   I don't see that mentioned anywhere on the description page.  In fact,
now that I look closer, I don't see any mention of whether this is
supposed to be a "same way" or "same thing" benchmark.  It used to be in
the "same way" category.
 
> Is this program and the Python program faster because it does fewer
> accesses to the collection? (Better algorithm not faster access).

The paper linked from the description page mentions changing the
algorithm in order to increase speed.  If that's not allowed, we should
probably mention it.

http://www.apl.jhu.edu/~hall/text/Papers/Lisp-Benchmarking-and-Fannkuch.ps

"The algorithm generates each of the n! permutations of n integers and
computes the number offlips. It would be better, of course, to generate
the permutations in a way that poorer ones can be immediately eliminated
from further consideration. For example, the heuristic, "The best
permutation can not start with 1 or end with n", leads to a simple check
that saves about nineteenpercent of the effort."


Greg Buchholz




		
__________________________________ 
Do you Yahoo!? 
Yahoo! Mail - 250MB free storage. Do more. Manage less. 
http://info.mail.yahoo.com/mail_250