Re: [gnugo-devel] Reading patch

[Paul Pogonyshev]

Nando wrote:
> I just finished testing the appended patch with following results which
> look rather good to me. The breakage :
>
> trevora:220    PASS
> trevorc:1440   FAIL   the test would pass with --owl-node-limit 1250
> global:4       FAIL   current CVS would also fail both tests if the owl
> node global:5       FAIL   limit would be similarly increased (1250)
> tactics1:101   PASS
> tactics1:104   PASS
>
> Performance impact :
>
> - reading nodes: +0.6%
> - owl nodes: +0.3%
> - connection nodes: -eps
>
>
> Globally, I think the patch is just logical and removes a rotational
> dependency.

it looks like what Evan once tried (and i also did later). though the
patch looks logical, it has strange outcomes in that it doesn't seem
to really improve anything. maybe the reason is in the way we use
attack_either() ?

> -    int libs[2];
> -    int alibs = findlib(astr, 2, libs);
> +    int alibs[2];
> +    int blibs[2];
> +    int alib = findlib(astr, 2, alibs);

we should really come up with some convention on array/number of
elements naming. e.g. we use `libs[]' and `liberties' in board.c, but
`adjs[]' and `neighbors' or `adjs[]' and `num_adj' or `adj[]' and
`adjs' in chainlinks handling code in reading.c.

Paul