Re: [gnugo-devel] connection reader speed optimization

[Paul Pogonyshev]

Arend wrote:
> Paul wrote:
> > This patch speeds connection reading up noticeably.  Unlike most my
> > other speed optimization patches, this one does not target a function
> > (spread_connection_distances) directly, but rather the functions it
> > calls in turn.
>
> One change in this patch is that only the origin, not all stones, of a
> string get added to the connection queue. Is this necessary/useful?
>
> I found it quite convenient in uses of the connection queue (e.g. in my
> escape owl patch) to have all stones explicitly in there. And already
> now you have two places in the code where you need to expand the queue
> element containing the origin to the complete list of stones in the
> string -- does that really make the code clearer?
> (Paul already noted that this also causes the bad interaction with the
> break-in code at least for ego:10 -- that could be worked around,
> however.)

Well, this was a speed optimization attempt :)  It reduces the number of
queue scans a lot, but on the second thought that probably doesn't matter
much in terms of speed.  So, this change can be taken out of the patch.

However, i have a counter-proposal.  Is it really the intention to have
partial strings in break-in goal array?  Maybe fixing this will give an
improvement in break-in efficiency?  In other words i propose you to
investigate the matter :)

Paul