[computer-go] Rotate the dog or the tail?

[Chrilly]

> Spelling error.  Decision tree.  Just a bunch of nested if statements.
>
I first thought that it had something to do with "deca", ten. Some very 
clever 10-way branching tree.

Thanks for the very interesting information. It is an advantage for 
potential competitors to have this information. But in first place you have 
slowed down the Suzie development. So far I was happy with my design, but 
its very tempting to do it in the Many-Faces and Handtalk way and to 
implement the move-pattern also as hardcoded switch-statements. The general 
pattern-matching approach seems cleaner, but the hardcoded approach is more 
flexible (e.g. one can easily check for additional conditions) and faster.

In Suzie I store only the color of the point and the info about the 
8-Adjacent Neigbours in a 32-Bits.

typedef union {
    Uint16 Adjacent16;
    struct {
        Uint8 wAdjacent;
        Uint8 bAdjacent;
    };
} Adjacent_t;

typedef struct {
    Adjacent_t wbAdjacent;
    Uint8         LibAdjacent;
    Uint8        Stone;
};

The union is only used because its sometimes handy to access the information 
about black and white at the same time.
So every stone nows about the state of its 8-Adjacent neighbours. The 
information is somewhat redundant, because one would get the Liberty-Pattern 
by oring black and white. But its handy to have it directly. And it saves no 
space, because the compiler has anyway to pad the struct to 4-Bytes.
I use this 3x3 Tiles directly for pattern matching. One can speed up this, 
by storing the OR of  5 Tiles, Center, N,E,S,W. The first match is done with 
the OR of the corresponding board-tiles. This gives in most patterns already 
in the first compare a mismatch.

Chrilly