[m-dev.] ICFP Contest: Any signs of life?

[Peter Moulder]

One possibility is for the computer to take an existing sequence of
instructions, and apply some random perturbations to the lengths of each
of the segments of same command.  I.e. do the equivalent of `uniq -c'
over the input, then modify the counts randomly, rejecting those that
result in crashes, and accepting those that result in shorter times.

I believe the above technique is independent of other optimizations, btw.

However, without having tried it, I'd guess that it won't make huge
improvements, because it doesn't use knowledge about how to achieve
objectives like not crashing, it just flails about randomly.

Also, I've just run

  tr . \\n < Een.trc|uniq -c|less

and I notice that in many places it spreads its `r' commands out over a
turn rather than bunching them together; I don't know whether or not this
is a common characteristic of optimal traces or whether it's just an
artifact of how that trace was generated.


Another thing I was considering is to modify a graphics program to handle
these funny curves, so that a human can do those perturbations, and apply
human intelligence to the problem.

pjm.

--------------------------------------------------------------------------
mercury-developers mailing list
Post messages to:       mercury-developers at cs.mu.oz.au
Administrative Queries: owner-mercury-developers at cs.mu.oz.au
Subscriptions:          mercury-developers-request at cs.mu.oz.au
--------------------------------------------------------------------------