Re: [gnugo-devel] dfa transformation optimization

[bump]

Paul wrote:

> at least owl_defendpat can be improved significantly (3-4 times)
> and this database is the one that does matter since it's the
> largest one.  when dfa optimization was first implemented, it
> gave about 2% speedup.  we the current state of .dtr files we
> can probably hope for more than 1%.
> 
> are we going to have a tuning period before 3.4?  it is better to
> do optimization after all dfa patterns are in.

Arend wrote:

> I think Paul's question was aimed at whether he should rebuild the .dtr
> files now -- if there are still owl tuning patches coming, then it is
> better to do that after they are in.

Maybe we should allow owl tuning patches up to 3.3.23 and rebuild 
the .dtr files just before that release (which could be in a week or two).

Dan