[computer-go] State of the art of pattern matching

[Mark Boon]

On 28-mrt-08, at 09:43, Jacques Basaldúa wrote:

> The first source code was just an example to see what kind of code
> is generated. The second is useful, if you understand asm you  
> should understand it.

Well, the only serious assembler I ever wrote was on a 6502 :-) And  
that was a very long time ago, so I'm sorry to say the asm is a bit  
too hard for me to see what it does. I suppose there's no reason why  
it has to be assembler, you could just as well generate C-code. Maybe  
C-compilers aren't good enough still compared to hand-crafted code in  
cases like these?

So although I start to understand some things, there's still a lot  
unclear to me. I think I see how you generate functions to  
efficiently update the hash-codes but I don't see yet how you go from  
there to finding patterns. I assume you allow some of the points to  
be undefined (also called 'don't care points') but I don't see how.  
And if you allow undefined points, why would you need masks of  
smaller manhatten distances?

Mark

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://computer-go.org/pipermail/computer-go/attachments/20080328/e4d78781/attachment.htm