[computer-go] Language War!

[Chrilly]

What is the advantage of a BitSet Representation over e.g. an Array of 
Bytes?
Board[point]=WHITE_STONE;  is faster than something link
WhiteBitSet[point>>5]|=(0x1<<(point&0x1F));

Same of course for the test if a stone is there.
One can calc with Bitsets the liberties by Bit-Shifting but it seems to be 
not trivial to calc the number of liberties for each string.
I have used Bitsets for a fast Influence function. But it did not really 
work, because this is a measure of the most nearby stone. But one has no 
information of the number of  influence stones.

In Suzie the Board is a struct { Uint8 Stone, Uint8 LibertyPattern, Uin8 
StonePatter[2]}; LibertyPattern and StonePattern contain the info about the 
8-adjacent Neigbours.
This is some effort when Setting a stone on the board, because one has also 
to change the Patterns of the 8 Neighbour-Squares. But the Info is handy for 
a lot of operations.
The really slow part is that I maintain also a list of liberties (and not 
just the liberty count) and a list of strings which share at least one 
liberty with the string and a list of opponent adjacent strings. This can be 
tricky when (several) strings are merged or in the undo-operation split up.

STL:
Normally a special purpose data-representation is faster than a general 
purpose one (it one takes advantage of the additional information). I know 
only the results of J.Bently in "Programming Pearls, 2nd. edition", where 
STL is about 5 times slower than a handcrafted sorting routine.

Chrilly