Re:Computer speed

[Compgo123@xxxxxxx]

>..... the best way is to use FPGAs.
Then what's ASCICs? Are they the same thing?

Darren Cook mentioned using 8 networked PCs. Is it possible to let each
computer runing a program on its own? When the main computer need some
specific info or task to be done, it issues something like a hardware
interrupt to a specific computer. In this case, multitasking and multithreads
would be needed. The advantage is that the computers can use the opponent's
time. For example, each computer can be in charge one specific group of
stones, calculating its properties (# of liberties, eyes, etc.) and ways to
extend that group. Or each computer can be in charge of a different task.

>..... It's easy to press 361 small processors
>onto a single go-cpu, each in-logic-go-unit already a hundred times 
>faster than the same program running at a PII cpu.

It's an interesting idea to using 361 processors to simulate a Go board. I
think the difficulty is how to connect all these processors in a meaningful
way. As one position is concerned, it very simple. It only need info about the
color of its stone and info about 4 neighbours. When stones are connected into
chains, then the property of the chain becomes important instead the
properties of each indivual position. As number of stones in the chain
increase, the number of properties of the chain also increases (in general,
but not always). For a two stone chain, properties are # of liberties, status
of 6 neighbours. For a three stone chain, one also need to specify the shape
of the chain. What are the topological properties of a n-stone chain?

However, if we only talk about topological properties of stones, we may miss
something very big. Often people say Go game is more complicated because it
has 3^361 end game variations. Well that's only the tip of an iceberg, because
there are about ((Nx/2)!)**2 ways to reach just one end position. Nx is the
total number of stones on the board at the end of the game. On average there
are more than 200 stones on the board at the game end.  Thus, the later number
is much larger than 3^361. This probably is the reason why many people are
puzzled by the question why Go is so difficult to program considering it's
much more (if not completly ) a mathematical game than chess. It's
mathematical alright. Unfortunately it's in a field that is not well charted.
What should we call it. Topological combinatorial?  Combinatorial topology?
Sequential topology ? Or Combinatorial sequential topology? And it's at a time
when mathematician is becoming an endangered species. It's no mystry why we
cannot find a effective search method. That's because we don't know what we
are looking for in a search in a Go game at the first place. It's like a guy
passing by asking 'Hey buddy, what are you looking for?'  The other guy
answers 'Well, I'll first find out how I'm going to find it. Then I'll find
out what I'm looking for.' (Well it may exagerate a liitle.)


Dan Liu