Re: computer-go: FPGA

[Vincent Diepeveen <diep@xxxxxxxxx>]

Even worse as this: my chessprograms evaluation doesn't fit into
FPGA very easily unless i get some real expensive (and slow) components.

Making it in ASIC is even harder. It won't fit easily on the chip.

Now go programs have the tendency to rely basically upon the knowledge
inside, so here you go... ...how do you tell a student that the next 10
years he'll be busy translating things to fpga?

Apart from that, it's outdated by then. See Deep Blue outdated search.
It searched 11 to 13 ply at 480 ASIC processors WITHOUT hashtables.
each ASIC processor did 6 ply in hardware. 

I get 11 to 13 ply Single CPU nowadays. with 40k nps a second, but WITH
shared memory. 

At 06:08 AM 8/29/00 -0700, you wrote:
>Another item to reduce costs if you wish to pursue this, at least
>on an initially limited basis... several universities have training in
>Verilog and VHDL, along with actual FPGA work, as part of
>their computer engineering programs (my direct experience
>being CSU Sacramento). If you talk with those in charge of the
>programs, you might even be able to convince some enterprising
>student in desperate need of a senior project to help. Otherwise,
>I hope you know the appropriate hardware definition languages. :)
>
>I think that the experiment might be worth the attempt to assist
>some areas of the program. However, that being said, I think
>David has a very key point that folks sometimes forget: you
>cannot efficiently fix slow code by throwing a faster piece of
>technology at it, despite what managers may say.
>
>Hmm... that's enough for a momentary delurking....
>
>  Kurt
>
>
>David Fotland wrote:
>
>> Many Faces of Go does about 30,000 tactical search nodes per second
>> on a Pentium-450, so there is no need for an FPGA.  Just write faster
>> code :)  MFGO's code is not particularly optimized for speed, since I
>> have a very big and complex move generator.  Other programs have much
>> faster local searching than Many Faces.
>>
>> And you can buy FPGA PCI cards for a few hundred dollars.  Do an
>> internet search to find where.
>>
>> -David Fotland
>>
>> At 10:02 PM 8/28/00 EDT, Compgo123@xxxxxxxxxxxxxxxxx wrote:
>> >Recently I tried to write code for local tactical search. To improve the
>> >search result, the code becomes more and more complicated and takes
more and
>> >more time. It now can only search about 20 end nodes per second on a
233 MHz
>> >machine and it still need more code to produce desired results. It
seems to
>> >me a hardware improvement is needed to produce a good program.
>> >
>>
>>  Then there is the cost to implement the
>> >FPGA in a PCI card. Many company will do this for you but with a big
charge
>> >(probably more than $10,000). The total cost to implement such a PCI card
>> >with FPGA could run as high as $20,000 to $30,000. Unless one can find
some
>> >funding, it's difficult for an individual to do.
>> >
>> >
>> >Dan Liu
>> >
>> >
>
>
>