Hi all,
well, the multi-flaring command-line options for fc-solve have been completely
implemented, and appear to be working nicely, though I discovered a few bugs
in the course of experimentation and if one opts to use them, they will need
to accommodate for that, thus possibly breaking some forwards-backwards-
compatibility (or something - ;-) ).
After all that worked, I started working on the
"fc-solve/presets/soft-threads/meta-moves/auto-gen" sub-directory in an
attempt to generate better presets. So I wrote some code to implement my idea
of an algorithm where quotas are gradually allocated to flares that together
yield the shortest solution on average. I noticed this had some local extrema,
and finally yielded results that were much worse than my previous proof-of-
concept 90% percentile flare based solver.
Eventually, I tried to reproduce the original scans exactly, and reached
roughly the same average. Which means that my optimisation algorithm yields
sub-optimal results (possibly due to the local extrema).
I've been thinking about some alternative, more analytical algorithms like
finding which scans yield the minimums for most of the boards, but it's not
done yet. There seems like there are many factors at play here due to the
solution length / iterations count dipole.
Regards,
Shlomi Fish
--
-----------------------------------------------------------------
Shlomi Fish http://www.shlomifish.org/
First stop for Perl beginners - http://perl-begin.org/
God considered inflicting XSLT as the tenth plague of Egypt, but then
decided against it because he thought it would be too evil.
Please reply to list if it's a mailing list post - http://shlom.in/reply .
Received on Sun May 16 2010 - 07:14:19 IDT