There is a fairly recent paper dealing with parallelization of UCT by Tristan
Cazenave and Nicolas Jouandeau.
On the parallelization of UCT.
In CGW 2007, pages 93-101, June 2007.
[ http://www.ai.univ-paris8.fr/~cazenave/parallelUCT.pdf ]
The authors tried three methods.
1) multiple UCT algorithms in parallel, no communication, master adds the
number of games and numbers of wins
2) multiple runs, some shared information used to build UCT trees in the slave
nodes.
3) parallelize at the leaves, heavy communications.
Their experiment used up to 16 machines connected via ethernet. The concluded
that speedup was good with modest communications,
but heavy communications in the 3rd model actually caused a 16-node machine to
be slower than 8 nodes.
With today's quad processors using shared memory,performance tradeoffs are
likely to differ.
Nonetheless, their paper offers a good starting point.
Terry McIntyre <[EMAIL PROTECTED]>
“Wherever is found what is called a paternal government, there is found state
education. It has been discovered that the best way to insure implicit
obedience is to commence tyranny in the nursery.”
Benjamin Disraeli, Speech in the House of Commons [June 15, 1874]
----- Original Message ----
From: Gian-Carlo Pascutto <[EMAIL PROTECTED]>
To: computer-go <[email protected]>
Sent: Saturday, January 12, 2008 2:30:17 PM
Subject: Re: [computer-go] handicap for mogo against pro ?
Olivier Teytaud wrote:
> Also, there are probably other people interested in combining
> UCT and mpi; I don't know if some people have a more precise idea
> of the level of the MPI+UCT combination than us. Someone ?
MPI is just a parallel programming model/library, right?
So the only thing to know is the effective[1] speedup of the MPI
version, and how well UCT scales with increasing timecontrols/speed.
I believe Don has data on the latter and you should have data on the
former.
[1] How much faster you find the correct move. Not interesting is: how
many positions you search per second or how many playouts you do per
second.
--
GCP
_______________________________________________
computer-go mailing list
[email protected]
http://www.computer-go.org/mailman/listinfo/computer-go/
____________________________________________________________________________________
Never miss a thing. Make Yahoo your home page.
http://www.yahoo.com/r/hs_______________________________________________
computer-go mailing list
[email protected]
http://www.computer-go.org/mailman/listinfo/computer-go/
