Re: [gnugo-devel] Patch: Caching for owl reading

[Arend Bayer]

Inge wrote:
> Arend wrote:
> > This patch converts the break-in code to use Inge's new cache. It will
> > conflict slightly with Inge's owl patch, as I had to change the prototyp
> > of tt_get, too (adding the additional hash value to identify the goal).
> > But it should be trivial to merge.
> >
> > No regression changes (*), almost no node count change.
>
> This doesn't surprise me in the least (the no node count change).  You see,
> Arend, that what you did was to search in the cache for reading results with
> a special hash value for the goal.  But you never stored it so there was
> nothing to find.  In fact, it surprises me that you didn't get more nodes.
> Perhaps break-in reading doesn't gain much by caching.

Ooops. I must simply have made the test run in the wrong directory.

Thanks for pointing this out.

Arend