Re: [computer-go] Hardware-Instruction.

[Vincent Diepeveen <diep@xxxxxxxxx>]

At 22:02 6-11-2004 +0100, chrilly wrote:
>
>-----Ursprüngliche Nachricht-----
>Von: Don Dailey <drd@xxxxxxxxxxxxxxxxx>
>An: computer-go@xxxxxxxxxxxxxxxxx <computer-go@xxxxxxxxxxxxxxxxx>
>Cc: computer-go@xxxxxxxxxxxxxxxxx <computer-go@xxxxxxxxxxxxxxxxx>
>Datum: Samstag, 6. November 2004 18:54
>Betreff: Re: [computer-go] Hardware-Instruction.
>
>
>>
>>> According my feeling is Cilk more high level than e.g. MPI. It is easier
>>> converting an existing programm to a parallel version with Cilk than with
>>> MPI. I assume (but have never done any concrete tests or studies) that
>this
>>> high-level comes with a price. MPI is probably faster and more flexible.
>>
>>Cilk is really C, so it's not a high level language.  Probably you are
>>right about  the flexibility issue,  Cilk does the  scheduling itself.
>
>This was my meaning of "high-level". In MPI the application (programmer) is
>responsible for scheduling.
>
>>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.
>At the Leiden 2004 tournament one of the running gags was that Diep has on
>the massive-parallel system too many Kilo-Bugs per second.
>When Fritz run in preperation for the Kramnik match the first time on an
>8-processor server it killed within a few milliseconds the OS.
>When Hydra (Brutus) was ported from a Dual to an 8-processor Cluster, it
>landed within 100 millisecond in a Deadlock.
>We could fix the problems within 1 week. Since then I have not touched the
>parallel code because it works fine. We had also no problem when going from
>8 to 16 processors. But for Parallel-Researchers is 16-processors not really
>parallel. Starts somewhere at 256-Processors.
>
>Question: Is Cilk available on Clusters?
>
>Chrilly

Yes there were already linux ports for it at around 1998 or so.

I am sure i no longer have it at my machines.

I must warn you. Cilk is fine for embarrassingly parallel jobs. Not for
search. Search means you non stop must start and stop processes. That is
already too expensive to do. Having global list of 'available workers' is
already too expensive. 

Cilk has another few work stealing things. They all feature lists of all
processors.

That is just too expensive. 

>
>
>_______________________________________________
>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/