CCRL Discussion Board

A while back I was reading about an egtb generator that mentioned the possibility of 7-men, but do to the intense memory usage it would require swapping to to disk. (Muller perhaps?).

Came across this today and thought maybe this would help with such a bottleneck.
http://www.physorg.com/news185438129.html#top

Essentially 2gig/s speeds on SSD.
Thoughts and comments are appreciated

-Josh

The problem with SSD drives is that the bytes on it are more expensive.
The advantage of mine and h.g.mullers algorithm is that they work well with normal hard disks because they don't need all that many seeks.
With mine, reading and writing disk in blocks of 16 MB is reasonable on a machine with 6-8 GB of ram.
As the results need to be stored anyway, you can use 6-8 disks in parallel and thus get speeds of perhaps 1 GB/s today without spending more than you have to.
(I am on the job and have hopes of getting the program written in 6 months.)

I haven't being paying much attention recently.

Have their been developments to EGT-algorithm algorithms that significantly reduce the number of I/O disc-transfers?

I conjecture that the best way to optimise is to notionally divide the index-range of positions into a number of subranges and 'park' requests for information about positions in those subranges until there is no alternative but to go and look up that subrange (which would be done in one disc transfer if the subrange was small enough or the disc transfer big enough).

The same idea works on a cluster-machine where a balance has to be struck between immediacy of request and congestion on the 'backplane' network.

g

I wrote an overview at: viewtopic.php?f=6&t=3652&p=47585#p47585
Then this thread has most explained: viewtopic.php?f=6&t=3458&p=34282#p34282