Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: SAN problem with Crafty's wall.pgn book

Author: Robert Hyatt

Date: 06:41:44 06/20/98

Go up one level in this thread


On June 19, 1998 at 17:41:59, Edward Screven wrote:

>On June 19, 1998 at 17:09:29, Robert Hyatt wrote:
>
>>This turns an O(N) operation (cpu complexity) and O(2N)  operation (Space)
>>into an O(N^2) (complexity) to get to O(N) space.  I don't like the trade
>>off.  I can build a 350,000 game book with 25,000,000 moves, in under 40
>>minutes on my P6, and under 10 minutes on a good alpha.  The factor of two
>>in space isn't worth the factor of N in speed, because N is *big*.
>
>O(N) --> O(N**2)?  no.  i'll assume by O(N) time you mean expected time
>given pgn input of reasonable size, since the sort is O(N log N), but
>the overall process is dominated by scanning the input for typical
>inputs.
>
>partitioning will go from O(N) --> O(N * K) time and O(N) --> O(N / K)
>space, where K is the constant number of partitions you choose.  you
>should choose K so that you just enough space to complete the operation.
>for example, if you require 500M of temporary space to build the book,
>but you only have 200M free, then you would choose K == 3, which would
>take about three times longer than if you had enough free disk space.
>
>whether or not it's worthwhile depends on what K you need to use and
>how badly you want the book.  when super-wall.pgn becomes available,
>wouldn't you be killing to run with K=3 or K=4 if it meant you could
>actually build a book out of it?
>
>  - edward

no...  because when N=60,000,000 multiple passes is simply gross in terms of
total time.  It's easier to go buy a 9 gig or a 23 gig SCSI disk, than to let
the book create run forever.  IE with 25,000,000 moves, my current book create
takes about 30 minutes.  Of that, 25 minutes + is spent on reading/parsing the
moves, the remainder is sorting/merging.

Sorting and merging times won't change, but the parsing time will be multiplied
by "K" which is a pretty large "product".  IE with K=.5 million, this takes
50 passes at 25 minutes per pass.  I'd rather deal with larger space
requirements than with 24 hours to build the book.



This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.