Re: [computer-go] Hardware-Instruction: Effort of parallization.

["chrilly" <chrilly@xxxxxxxx>]

>
>At the last world championship we played in, David Levy predicted that
>we  would not  win even  though  we were  one of  the favorites.   His
>prediction and reasoning was based  on the fact that parallel programs
>are  typically full of  bugs and  problems.
As described in my previous mail was the parallelization of Brutus/Hydra
with MPI relative smooth. The reason for this was Ulf Lorenz. He had written
with PConners a massive parellel programm before. Ulfs research topic is
parallel algorithms. I had no parallel experience and I would have been
simply lost. It is unbelievable what can go wrong in a parallel programm (on
a Dual almost nothing goes wrong, but live becomes interesting from 4
processors upwards). Personnally I would have either choosen Cilk or an
algorithm which uses no explicit synchronization. The processes just share
some hashtable information. But this loosly coupled algorithms are only
usefull for a Dual or at most a 4 processor version.

Chrilly



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