Re: [computer-go] SlugGo approach: GNU vs.Goliath

[Vincent Diepeveen <diep@xxxxxxxxx>]

At 16:11 6-1-2005 +0100, Marco Scheurer wrote:
>>> It is relevant. It's 2 times slower at least.
>> Depends on the platform, IMHO. for platforms with separate
>> floating point boxes ('units') such as the alpha (maybe newer pentiums 
>> ?)
>
>Apple's G4 and G5 have a really nice vector unit.

What's the price of a quad g5?

Quad opteron 2.2Ghz is selling for $9995 now and cheapest price at ebay i
see now is about $9500 for a complete working quad 2.2Ghz.

>> the float ops *could* be done in parallel.

>Yes, but the same thing would be true for ints.

G5 is very pathetic for integer code.

>> It's just an engineering decision, I guess.
>
>I'm all for floats. Why are high end GPU (for instance from nvidia) and 
>image file formats (for instance OpenEXR) starting to use floats?

Why are those GPU's clocked so low? CPU's clock already at 3.8Ghz now, i
can't remember any GPU clock that high. In fact does one well sold one
already clock above 1Ghz?

Note that GPU's are vector processors and that evaluation functions aren't
allowing in software to be vector processed.

Score of gameplaying evaluations can be estimated easily between say -1
million and +1 million, which fits real easy in a 32 bits integer, 
objects in open-gl can't.

Evaluation scores go with 3 instructions per cycle and floating point
doesn't. There is a hard limit of 1 instruction per cycle at PC processors
there.

If you want a hardware go engine, go talk to Chrilly Donninger.

Vincent

>marco
>
>_______________________________________________
>computer-go mailing list
>computer-go@xxxxxxxxxxxxxxxxx
>http://www.computer-go.org/mailman/listinfo/computer-go/
>
>
_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/